Upload 6 files

README.md

@@ -1,19 +1,49 @@
 ---
-title: FatigueAnalyzer
-emoji: 🚀
-colorFrom: red
+title: MineVision AI - Advanced Fatigue Analytics
+emoji: ⛏️
+colorFrom: blue
 colorTo: red
-sdk: docker
-app_port: 8501
-tags:
-- streamlit
+sdk: streamlit
+sdk_version: 1.38.0  # Ganti dengan versi streamlit yang digunakan
+app_file: app.py
 pinned: false
-short_description: Track. Analyze. Prevent fatigue.
+license: apache-2.0
 ---
 
-# Welcome to Streamlit!
+# MineVision AI - Advanced Fatigue Analytics
 
-Edit `/src/streamlit_app.py` to customize this app to your heart's desire. :heart:
+## Deskripsi
+Aplikasi ini adalah dashboard analitik kelelahan berbasis web yang dirancang untuk operasi pertambangan. Menggunakan data dari sistem deteksi kelelahan (seperti Wenco DSS), aplikasi ini menyediakan wawasan dan analisis real-time untuk membantu mengidentifikasi, menilai, dan mengelola risiko kelelahan operator. Tujuannya adalah untuk meningkatkan keselamatan kerja dan produktivitas dengan mengurangi kecelakaan yang terkait dengan kelelahan.
 
-If you have any questions, checkout our [documentation](https://docs.streamlit.io) and [community
-forums](https://discuss.streamlit.io).
+## Fitur Utama
+*   **Dashboard Eksekutif**: Menampilkan metrik keselamatan utama seperti total alert, jumlah operator dan aset, serta durasi rata-rata kejadian.
+*   **Analisis Tren**: Visualisasi tren kelelahan berdasarkan jam, shift, hari dalam seminggu, dan minggu.
+*   **Analisis Lanjutan**: Analisis berdasarkan jenis armada, kecepatan vs jam, durasi vs jam, distribusi kecepatan, dan distribusi operator per shift.
+*   **Kategorisasi Risiko Kelelahan**: Menganalisis kejadian berdasarkan matriks risiko kelelahan (Kritis, Tinggi, Sedang, Rendah) berdasarkan kecepatan dan waktu.
+*   **Wawasan Berbasis AI**: Ringkasan otomatis dan wawasan berdasarkan data yang dianalisis.
+*   **Asisten AI Interaktif**: Chatbot sederhana untuk menanyakan informasi tentang data kelelahan (operator terbanyak, shift terbanyak, dll.).
+
+## Teknologi yang Digunakan
+*   **Streamlit**: Framework untuk membuat aplikasi web interaktif dalam Python.
+*   **Pandas**: Manipulasi dan analisis data.
+*   **Plotly/Plotly Express**: Visualisasi data interaktif.
+*   **Openpyxl**: Pembacaan file Excel.
+
+## Cara Menggunakan
+1.  Akses aplikasi melalui URL Hugging Face Spaces.
+2.  Gunakan filter di sidebar untuk menyaring data berdasarkan Tahun, Bulan, Minggu, Rentang Tanggal, Operator, Shift, dan Rentang Jam.
+3.  Jelajahi berbagai bagian dashboard untuk memahami pola kelelahan.
+4.  Gunakan kotak chat "MineVision AI Assistant" di bagian atas untuk menanyakan pertanyaan spesifik tentang data.
+
+## Struktur Proyek
+*   `app.py`: File utama yang berisi kode aplikasi Streamlit.
+*   `requirements.txt`: File yang berisi daftar dependensi Python yang diperlukan untuk menjalankan aplikasi.
+*   `manual fatique.xlsx`: File data input contoh (jika disertakan dalam repositori).
+
+## Catatan
+*   Aplikasi ini dirancang untuk menganalisis data kelelahan operator dari file Excel. Pastikan struktur data masukan sesuai atau sesuaikan kode untuk membaca data dari sumber lain.
+*   Wawasan dan rekomendasi didasarkan pada analisis data historis dan prinsip-prinsip manajemen risiko kelelahan (FRMS).
+*   Asisten AI saat ini menyediakan jawaban berbasis aturan sederhana berdasarkan data yang tersedia dan informasi umum tentang FRMS. Ini bukan model AI canggih seperti GPT.
+
+## Lisensi
+Apache 2.0

app.py

@@ -0,0 +1,1791 @@
+import streamlit as st
+import pandas as pd
+import plotly.express as px
+import plotly.graph_objects as go
+from datetime import datetime, timedelta
+import requests
+import json
+import numpy as np
+import math
+import base64
+
+# =================== CONFIG =====================
+st.set_page_config(
+    page_title="MineVision AI - Advanced Fatigue Analytics",
+    page_icon="🛡️",  # Safety icon
+    layout="wide",
+    initial_sidebar_state="expanded"
+)
+
+# =================== LOGO =====================
+logo_path = "btech.png"  # File logo
+def get_base64(file_path):
+    with open(file_path, "rb") as f:
+        data = f.read()
+    return base64.b64encode(data).decode()
+
+try:
+    logo_base64 = get_base64(logo_path)
+    logo_html = f'<img src="data:image/png;base64,{logo_base64}" style="max-height: 80px; max-width: 120px;">'
+except FileNotFoundError:
+    st.warning(f"Logo file '{logo_path}' not found. Using placeholder text.")
+    logo_html = '<div style="font-size: 18px; font-weight: bold; color: #2c3e50;">BTECH</div>'
+
+# =================== GLOBAL CSS =====================
+st.markdown("""
+<style>
+body {
+    background-color: #f6f8fa;
+}
+
+/* ===== HEADER WRAPPER ===== */
+.header-container {
+    display: flex;
+    justify-content: space-between;
+    align-items: center;
+    padding: 25px 35px;
+    background: white; /* Latar belakang utama diubah menjadi putih */
+    border-radius: 0 0 14px 14px; /* Rounded bottom only */
+    box-shadow: 0 5px 18px rgba(0,0,0,0.15); /* Bayangan lebih lembut */
+    border: 1px solid #e0e0e0; /* Border tipis untuk definisi */
+    margin-bottom: 25px;
+    position: relative;
+    overflow: hidden; /* Ensure rounded corners clip content */
+}
+
+/* Optional: Subtle pattern or texture overlay (optional, can be removed) */
+/* .header-container::before {
+    content: "";
+    position: absolute;
+    top: 0;
+    left: 0;
+    right: 0;
+    bottom: 0;
+    background: linear-gradient(45deg, rgba(255,255,255,0.03) 25%, transparent 25%, transparent 50%, rgba(255,255,255,0.03) 50%, rgba(255,255,255,0.03) 75%, transparent 75%, transparent);
+    background-size: 20px 20px;
+    pointer-events: none;
+} */
+
+/* ===== HEADER TEXT ===== */
+.header-title {
+    color: #2c3e50; /* Teks header diubah agar kontras dengan latar putih */
+    font-family: 'Segoe UI', sans-serif;
+    flex-grow: 1; /* Allow text to take up available space */
+    margin-right: 20px; /* Space between text and logo */
+    text-align: left;
+}
+
+.header-title h1 {
+    font-size: 2.7em;
+    font-weight: 650;
+    margin: 0;
+    text-shadow: 1px 1px 2px rgba(0,0,0,0.1); /* Bayangan teks lebih lembut */
+}
+
+.header-title p {
+    font-size: 1.25em;
+    opacity: 0.85; /* Sedikit transparan untuk subjudul */
+    margin-top: 6px;
+    font-style: italic;
+    color: #34495e; /* Warna subjudul disesuaikan */
+}
+
+/* ===== LOGO WRAPPER ===== */
+.header-logo {
+    display: flex;
+    align-items: center;
+    justify-content: flex-end; /* Align logo to the right within its container */
+    flex-shrink: 0; /* Prevent logo container from shrinking */
+}
+
+/* ===== LOGO STYLE ===== */
+.header-logo img {
+    border-radius: 10px;
+    border: 2px solid rgba(44, 62, 80, 0.15); /* Border logo disesuaikan */
+    box-shadow: 0 3px 10px rgba(0,0,0,0.1); /* Bayangan logo lebih lembut */
+    max-height: 80px; /* Set max height */
+    max-width: 120px; /* Set max width */
+}
+
+/* ===== METRIC CARDS ===== */
+.metric-card {
+    background: #ffffff;
+    padding: 18px 22px;
+    border-radius: 12px;
+    border-left: 6px solid #1e3c72;
+    box-shadow: 0 3px 8px rgba(0,0,0,0.10);
+    transition: 0.25s ease-in-out;
+}
+.metric-card:hover {
+    transform: translateY(-4px);
+    box-shadow: 0 6px 15px rgba(0,0,0,0.18);
+}
+
+/* ===== INSIGHT BOX ===== */
+.insight-box {
+    background: #fafafa;
+    padding: 18px;
+    border-radius: 12px;
+    border-left: 6px solid #ff6b6b;
+    margin: 15px 0;
+    box-shadow: 0 2px 6px rgba(0,0,0,0.08);
+}
+
+/* ===== RISK MATRIX ===== */
+.risk-matrix {
+    border-collapse: collapse;
+    width: 100%;
+    margin: 20px 0;
+}
+.risk-matrix th, .risk-matrix td {
+    border: 1px solid #ddd;
+    padding: 12px;
+    text-align: center;
+}
+.risk-matrix th {
+    background-color: #f2f2f2;
+}
+.critical { background-color: #ffcccc; font-weight: bold; }
+.high { background-color: #ffebcc; }
+.medium { background-color: #ffffcc; }
+.low { background-color: #e6ffe6; }
+
+/* ===== CHAT UI ===== */
+.chat-container {
+    background: white;
+    padding: 20px;
+    border-radius: 12px;
+    height: 400px;
+    overflow-y: auto;
+    border: 1px solid #ccc;
+}
+.user-message {
+    background: #e3f2fd;
+    color: black;
+    padding: 12px;
+    border-radius: 12px;
+    margin: 10px 0;
+    text-align: right;
+    border: 1px solid #bbdefb;
+}
+.ai-message {
+    background: #f5f5f5;
+    color: black;
+    padding: 12px;
+    border-radius: 12px;
+    margin: 10px 0;
+    text-align: left;
+    border: 1px solid #e0e0e0;
+}
+
+/* ===== INPUT BOX ===== */
+.chat-box, .user-question, .ai-answer {
+    background: white;
+    border: 1px solid #ccc;
+    border-radius: 10px;
+    padding: 12px;
+    margin-bottom: 12px;
+}
+
+/* ===== FOOTER ===== */
+.footer {
+    text-align: center;
+    padding: 20px;
+    color: gray;
+    font-size: 0.9em;
+}
+
+/* ===== HOVER EFFECTS ===== */
+.metric-card:hover, .insight-box:hover {
+    box-shadow: 0 6px 15px rgba(0,0,0,0.2);
+    transition: all 0.3s ease-in-out;
+}
+
+</style>
+""", unsafe_allow_html=True)
+
+# =================== HEADER =====================
+st.markdown(f"""
+<div class="header-container">
+    <div class="header-title">
+        <h1>Safety Analysis and AI - Advanced Fatigue Analysis</h1>
+        <p>Proactive Safety Intelligence for Mining Operations</p>
+    </div>
+    <div class="header-logo">
+        {logo_html}
+    </div>
+</div>
+""", unsafe_allow_html=True)
+
+# Sisa kode Anda (LOAD DATA, FILTERS, VISUALISASI, dll.) tetap sama di bawah ini
+# ... (Kode selanjutnya disalin dari bagian bawah file Anda, misalnya LOAD DATA ke bawah)
+# =================== LOAD DATA ======================
+@st.cache_data
+def load_data():
+    try:
+        # ==================================
+        # 1. LOAD CSV & NORMALIZE COLUMNS
+        # ==================================
+        df = pd.read_csv("data.csv")
+        original_columns = df.columns.tolist()
+        # Normalize: lower, strip, underscore
+        df.columns = (
+            df.columns.astype(str)
+            .str.strip()
+            .str.lower()
+            .str.replace(r"\s+", "_", regex=True)
+        )
+
+        # ==================================
+        # 2. AUTO-DETECT COLUMNS (case-insensitive)
+        # ==================================
+        col_operator = next((c for c in df.columns if "operator" in c or "driver" in c), None)
+        col_shift = next((c for c in df.columns if "shift" in c), None)
+        # ✅ FIX: Search for normalized "parent_fleet", NOT original "Parent Fleet"
+        col_fleet_type = next((c for c in df.columns if "parent_fleet" in c), None)
+        col_fleet_no = next((c for c in df.columns if "fleet_number" in c), None)
+
+        # ==================================
+        # 3. DERIVE COLUMNS
+        # ==================================
+        # Unit Number
+        if col_fleet_no:
+            df["unit_no"] = df[col_fleet_no].astype(str).str.split("-", n=1).str[-1].str.strip()
+        else:
+            df["unit_no"] = "UNKNOWN"
+
+        # Speed
+        col_speed = None
+        for orig in original_columns:
+            norm = orig.lower().replace(" ", "_")
+            if "(in_km/hour).1" in norm or "speed" in norm:
+                if norm in df.columns:
+                    col_speed = norm
+                    break
+        if not col_speed:
+            col_speed = next((c for c in df.columns if "speed" in c), None)
+
+        # Time
+        time_cols = [c for c in df.columns if "gmt" in c and "wita" in c]
+        if len(time_cols) >= 2:
+            df["start"] = pd.to_datetime(df[time_cols[0]], errors="coerce")
+            df["end"] = pd.to_datetime(df[time_cols[1]], errors="coerce")
+        elif len(time_cols) == 1:
+            df["start"] = pd.to_datetime(df[time_cols[0]], errors="coerce")
+            df["end"] = df["start"] + pd.Timedelta(minutes=1)
+        else:
+            df["start"] = pd.NaT
+            df["end"] = pd.NaT
+
+        # Time features
+        if not df["start"].isna().all():
+            df["hour"] = df["start"].dt.hour
+            df["date"] = df["start"].dt.date
+            df["day_of_week"] = df["start"].dt.day_name()
+            # df["week"], df["month"], df["year"] — optional, not used in filters
+        else:
+            df["hour"] = 0
+            df["date"] = None
+
+        # Shift as int
+        if col_shift:
+            df[col_shift] = pd.to_numeric(df[col_shift], errors="coerce").astype("Int64")
+
+        # ✅ FIX: CREATE site & group_model HERE (not in sidebar!)
+        if col_fleet_type:
+            # Split ONCE on first '-', keep FULL left part (e.g., "Amsterdam - CAT789" → "AMSTERDAM")
+            split = df[col_fleet_type].astype(str).str.split("-", n=1, expand=True)
+            df["site"] = split[0].str.strip().str.upper()
+            df["group_model"] = split[1].str.strip().fillna("UNKNOWN").replace("", "UNKNOWN")
+        else:
+            df["site"] = "UNKNOWN"
+            df["group_model"] = "UNKNOWN"
+
+        return df, col_operator, col_shift, col_fleet_type, col_speed, col_fleet_no
+
+    except Exception as e:
+        st.error(f"Error loading data: {e}")
+        return pd.DataFrame(), None, None, None, None, None
+
+# ==================================
+# CALL load_data()
+# ==================================
+df, col_operator, col_shift, col_fleet_type, col_speed, col_fleet_no = load_data()
+df_original_full = df.copy()
+if df.empty:
+    st.stop()
+st.success("Data Loaded Successfully")
+df_full_report = df.copy()
+
+# =================== FILTERS (Sidebar) =====================
+filter_dict = {}
+
+with st.sidebar.form("filters_form"):
+    # ---------------- Date Range ----------------
+    if 'date' in df.columns and not df['date'].isna().all():
+        min_date = pd.to_datetime(df['date']).min().date()
+        max_date = pd.to_datetime(df['date']).max().date()
+        date_range = st.date_input("Select Date Range", (min_date, max_date))
+        filter_dict['date_range'] = date_range
+    else:
+        filter_dict['date_range'] = (None, None)
+
+    # ✅ FIXED: Use df['site'] & df['group_model'] (already created in load_data)
+    # ---------------- Site Filter ----------------
+    all_sites = sorted(df['site'].dropna().unique())
+    selected_site = st.selectbox(
+        "Filter Site",
+        options=[None] + all_sites,
+        format_func=lambda x: "All" if x is None else x
+    )
+    filter_dict['site'] = selected_site
+
+    # ---------------- Group Model Filter ✅ NOW WORKING ----------------
+    all_models = sorted(df['group_model'].dropna().unique())
+    selected_model = st.selectbox(
+        "Filter Group Model",
+        options=[None] + all_models,
+        format_func=lambda x: "All" if x is None else x
+    )
+    filter_dict['group_model'] = selected_model
+
+    # ---------------- Shift ----------------
+    if col_shift:
+        shifts = sorted(df[col_shift].dropna().unique())
+        selected_shift = st.selectbox(
+            f"Select {col_shift.replace('_', ' ').title()}",
+            options=[None] + shifts,
+            format_func=lambda x: "All" if x is None else f"Shift {x}"
+        )
+        filter_dict['shift'] = selected_shift
+    else:
+        filter_dict['shift'] = None
+
+    # ---------------- Operator ----------------
+    if col_operator:
+        ops = sorted(df[col_operator].dropna().unique())
+        selected_op = st.selectbox(
+            f"Select {col_operator.replace('_', ' ').title()}",
+            options=[None] + ops,
+            format_func=lambda x: "All" if x is None else x
+        )
+        filter_dict['operator'] = selected_op
+    else:
+        filter_dict['operator'] = None
+
+    # ---------------- Hour ----------------
+    if 'hour' in df.columns and not df['hour'].isna().all():
+        hours = sorted(df['hour'].dropna().unique())
+        hour_range = st.slider("Select Hour Range", int(min(hours)), int(max(hours)), (int(min(hours)), int(max(hours))))
+        filter_dict['hour_range'] = hour_range
+    else:
+        filter_dict['hour_range'] = (0, 23)
+
+    # ---------------- Unit No ----------------
+    if 'unit_no' in df.columns:
+        units = sorted(df['unit_no'].dropna().unique())
+        selected_unit = st.selectbox("Select Unit Number", [None] + units, format_func=lambda x: "All" if x is None else x)
+        filter_dict['unit_no'] = selected_unit
+    else:
+        filter_dict['unit_no'] = None
+
+    # ---------------- Submit ----------------
+    apply_filters = st.form_submit_button("Apply Filters")
+# =================== APPLY FILTERS =====================
+if apply_filters:
+    # Filter Date Range
+    if filter_dict.get('date_range'):
+        start_date, end_date = filter_dict['date_range']
+        df = df[(df['date'] >= start_date) & (df['date'] <= end_date)]
+
+    # Filter Site
+    if filter_dict.get('site') is not None:
+        df = df[df['site'] == filter_dict['site']]
+
+    # Filter Group Model
+    if filter_dict.get('group_model') is not None:
+        df = df[df['group_model'] == filter_dict['group_model']]
+
+    # Filter Shift
+    if filter_dict.get('shift') is not None:
+        df = df[df[col_shift] == filter_dict['shift']]
+
+    # Filter Operator
+    if filter_dict.get('operator') is not None:
+        df = df[df[col_operator] == filter_dict['operator']]
+
+    # Filter Hour Range
+    if filter_dict.get('hour_range'):
+        hr_start, hr_end = filter_dict['hour_range']
+        df = df[(df['hour'] >= hr_start) & (df['hour'] <= hr_end)]
+
+    # Filter Unit No
+    if filter_dict.get('unit_no') is not None:
+        df = df[df[col_fleet_no] == filter_dict['unit_no']]
+
+
+# Sisanya dari kode Anda (Visualisasi, dll.) tetap sama
+# Objective 1
+# ===================== GLOBAL FUNCTION: Hour Category Labels =====================
+def hour_range_label_full(hour):
+    if not (0 <= hour < 24):
+        return 'Unknown'
+    if 6 <= hour < 9:
+        return 'Shift 1 Morning Early (6-9)'
+    elif 9 <= hour < 12:
+        return 'Shift 1 Morning Late (9-12)'
+    elif 12 <= hour < 15:
+        return 'Shift 1 Afternoon Early (12-15)'
+    elif 15 <= hour < 18:
+        return 'Shift 1 Afternoon Late (15-18)'
+    elif 18 <= hour < 21:
+        return 'Shift 2 Evening Early (18-21)'
+    elif 21 <= hour < 24:
+        return 'Shift 2 Evening Late (21-24)'
+    elif 0 <= hour < 3:
+        return 'Shift 2 Dawn Early (0-3)'
+    elif 3 <= hour < 6:
+        return 'Shift 2 Dawn Late (3-6)'
+    return 'Unknown'
+
+# ===================== MAIN VISUALIZATION =====================
+st.subheader("Objective 1: Visualizing Operator Fatigue Patterns by Shift Hours")
+if 'start' in df.columns and not df.empty:
+    try:
+        # --- Data Preparation ---
+        df_local = df.copy()
+        if not pd.api.types.is_datetime64_any_dtype(df_local['start']):
+            df_local['start'] = pd.to_datetime(df_local['start'], errors='coerce')
+        df_local = df_local.dropna(subset=['start'])
+        df_local['hour'] = df_local['start'].dt.hour
+        # --- COLOR MAP: KUNING-ORANGE (Shift 1), BIRU (Shift 2) ---
+        color_map_full = {
+            'Shift 1 Morning Early (6-9)':      '#FFEB3B',  # Yellow 300
+            'Shift 1 Morning Late (9-12)':      '#FFC107',  # Amber 300
+            'Shift 1 Afternoon Early (12-15)':  '#FF9800',  # Orange 300
+            'Shift 1 Afternoon Late (15-18)':   '#F57C00',  # Deep Orange 300
+            'Shift 2 Evening Early (18-21)':    '#42A5F5',  # Light Blue 300
+            'Shift 2 Evening Late (21-24)':     '#1976D2',  # Blue 300
+            'Shift 2 Dawn Early (0-3)':         '#0288D1',  # Cyan 300
+            'Shift 2 Dawn Late (3-6)':          '#01579B',  # Blue 800
+        }
+        # --- Define intervals in analog-clock order (12→3→6→9) ---
+        intervals_shift1 = [(12, 15), (15, 18), (6, 9), (9, 12)]
+        labels_shift1 = [
+            'Shift 1 Afternoon Early (12-15)',
+            'Shift 1 Afternoon Late (15-18)',
+            'Shift 1 Morning Early (6-9)',
+            'Shift 1 Morning Late (9-12)',
+        ]
+        intervals_shift2 = [(0, 3), (3, 6), (18, 21), (21, 24)]
+        labels_shift2 = [
+            'Shift 2 Dawn Early (0-3)',
+            'Shift 2 Dawn Late (3-6)',
+            'Shift 2 Evening Early (18-21)',
+            'Shift 2 Evening Late (21-24)',
+        ]
+        # --- Compute frequencies ---
+        def compute_counts(intervals):
+            counts = []
+            for start_h, end_h in intervals:
+                cnt = df_local[(df_local['hour'] >= start_h) & (df_local['hour'] < end_h)].shape[0]
+                counts.append(cnt)
+            return counts
+        freq_shift1 = compute_counts(intervals_shift1)
+        freq_shift2 = compute_counts(intervals_shift2)
+        # --- Polar geometry ---
+        theta_midpoints = [45, 135, 225, 315]  # centers of 90° segments
+        bar_width = [90] * 4
+        angular_tick_vals = [0, 90, 180, 270]  # fixed angle positions
+        # ✅ CUSTOM TICK LABELS PER SHIFT (sesuai permintaan Anda)
+        angular_tick_text_shift1 = ["12", "15", "6/18", "9"]   # 0°, 90°, 180°, 270°
+        angular_tick_text_shift2 = ["24", "3", "18/6", "21"]   # 0°=24, 90°=3, 180°=6, 270°=21
+        # --- Independent radial scales ---
+        max_r1 = max(freq_shift1) if freq_shift1 and max(freq_shift1) > 0 else 1
+        max_r2 = max(freq_shift2) if freq_shift2 and max(freq_shift2) > 0 else 1
+        # ============== FIGURE: SHIFT 1 (KUNING-ORANGE) ==============
+        fig1 = go.Figure()
+        fig1.add_trace(go.Barpolar(
+            r=freq_shift1,
+            theta=theta_midpoints,
+            width=bar_width,
+            marker_color=[color_map_full.get(lbl, '#FFEB3B') for lbl in labels_shift1],
+            marker_line_color="black",
+            marker_line_width=1.5,
+            opacity=0.93,
+            hovertemplate="<b>%{text}</b><br>Fatigue Incidents: %{r}<extra></extra>",
+            text=labels_shift1,
+        ))
+        fig1.update_layout(
+            title=dict(text="Shift 1 (06:00–18:00)", font=dict(size=18, color="#FF9800", family="Segoe UI")),
+            polar=dict(
+                bgcolor="rgba(255,248,225,0.7)",
+                angularaxis=dict(
+                    rotation=90,  # 12 at top
+                    direction="clockwise",
+                    tickmode='array',
+                    tickvals=angular_tick_vals,
+                    ticktext=angular_tick_text_shift1,  # ✅ 12, 15, 6, 9
+                    tickfont=dict(size=14, color="#5D4037", weight="bold"),
+                    showline=True,
+                    linewidth=1.2,
+                    linecolor="#FFD54F",
+                ),
+                radialaxis=dict(
+                    visible=True,
+                    showticklabels=True,
+                    tickfont=dict(size=11),
+                    angle=45,
+                    gridcolor="#FFE082",
+                    gridwidth=0.8,
+                    range=[0, max_r1 * 1.15],
+                )
+            ),
+            showlegend=False,
+            height=550,
+            width=550,
+            margin=dict(t=65, b=40, l=40, r=40),
+            font=dict(family="Segoe UI, -apple-system, sans-serif"),
+        )
+        # ============== FIGURE: SHIFT 2 (BIRU) ==============
+        fig2 = go.Figure()
+        fig2.add_trace(go.Barpolar(
+            r=freq_shift2,
+            theta=theta_midpoints,
+            width=bar_width,
+            marker_color=[color_map_full.get(lbl, '#42A5F5') for lbl in labels_shift2],
+            marker_line_color="black",
+            marker_line_width=1.5,
+            opacity=0.93,
+            hovertemplate="<b>%{text}</b><br>Fatigue Incidents: %{r}<extra></extra>",
+            text=labels_shift2,
+        ))
+        fig2.update_layout(
+            title=dict(text="Shift 2 (18:00–06:00)", font=dict(size=18, color="#1976D2", family="Segoe UI")),
+            polar=dict(
+                bgcolor="rgba(230,245,255,0.7)",
+                angularaxis=dict(
+                    rotation=90,
+                    direction="clockwise",
+                    tickmode='array',
+                    tickvals=angular_tick_vals,
+                    ticktext=angular_tick_text_shift2,  # ✅ 24, 3, 6, 21
+                    tickfont=dict(size=14, color="#0D47A1", weight="bold"),
+                    showline=True,
+                    linewidth=1.2,
+                    linecolor="#64B5F6",
+                ),
+                radialaxis=dict(
+                    visible=True,
+                    showticklabels=True,
+                    tickfont=dict(size=11),
+                    angle=45,
+                    gridcolor="#BBDEFB",
+                    gridwidth=0.8,
+                    range=[0, max_r2 * 1.15],  # ✅ SKALA INDEPENDEN
+                )
+            ),
+            showlegend=False,
+            height=550,
+            width=550,
+            margin=dict(t=65, b=40, l=40, r=40),
+            font=dict(family="Segoe UI, -apple-system, sans-serif"),
+        )
+        # ============== EXPLANATION — URUTAN KRONOLOGIS, REMARK TETAP ==============
+        st.markdown("""
+        <div style="
+            background: linear-gradient(135deg, #FFFDE7 0%, #E3F2FD 100%);
+            padding: 20px;
+            border-radius: 12px;
+            border-left: 5px solid #FF9800;
+            margin: 22px 0;
+            box-shadow: 0 3px 10px rgba(0,0,0,0.06);
+        ">
+        <h4 style="color:#1976D2; margin:0 0 14px 0; display:flex; align-items:center;">
+            <span style="background:#FF9800; color:white; width:26px; height:26px; border-radius:50%; 
+                        display:inline-flex; align-items:center; justify-content:center; margin-right:10px; font-weight:bold;">!</span>
+            ⚠️ Clockwise Time Mapping (Analog Layout)
+        </h4>
+        <table style="width:100%; font-size:14px; border-collapse:collapse; color:#424242;">
+          <tr style="background-color:#FFF8E1;">
+            <th style="padding:8px; text-align:left; width:25%;">Time Block</th>
+            <th style="padding:8px; text-align:left;">Shift 1 (Day)</th>
+            <th style="padding:8px; text-align:left;">Shift 2 (Night)</th>
+          </tr>
+          <tr>
+            <td style="padding:8px; font-weight:bold;">1st Block</td>
+            <td><b>06 → 09</b></td>
+            <td><b>18 → 21</b> (Shift Start)</td>
+          </tr>
+          <tr style="background-color:#F5F9FF;">
+            <td style="padding:8px; font-weight:bold;">2nd Block</td>
+            <td><b>09 → 12</b></td>
+            <td><b>21 → 24</b> (Alertness Decline)</td>
+          </tr>
+          <tr>
+            <td style="padding:8px; font-weight:bold;">3rd Block</td>
+            <td><b>12 → 15</b></td>
+            <td><b>24 → 03</b> (Circadian Nadir)</td>
+          </tr>
+          <tr style="background-color:#F5F9FF;">
+            <td style="padding:8px; font-weight:bold;">4th Block</td>
+            <td><b>15 → 18</b></td>
+            <td><b>03 → 06</b></td>
+          </tr>
+        </table>
+        <p style="margin-top:12px; font-size:13px; color:#546E7A;">
+            <b>Scale is independent per shift</b> — bar length shows relative risk <i>within</i> the shift.
+        </p>
+        </div>
+        """, unsafe_allow_html=True)
+        # ============== RENDER CHARTS HORIZONTALLY (NO OVERLAP) ==============
+        col1, col2 = st.columns(2)
+        with col1:
+            st.plotly_chart(fig1, use_container_width=True, config={'displayModeBar': False})
+        with col2:
+            st.plotly_chart(fig2, use_container_width=True, config={'displayModeBar': False})
+        # ============== FOOTNOTE (SEMINAR-READY) ==============
+        st.caption(
+            " *Safety Insight*: Highest fatigue risk occurs during **24→06** (Shift 2) — aligns with circadian trough (Czeisler, 1999). "
+        )
+    except Exception as e:
+        st.error(f"⚠️ Rendering error: {e}")
+        st.code(f"{type(e).__name__}: {str(e)}", language="python")
+else:
+    st.info("⏳ Awaiting data... Ensure column `'start'` contains valid timestamps (e.g., '2025-06-15 14:30:00').")
+
+#Objective 
+#
+#Objective 1
+st.subheader("OBJECTIVE 2: Identify Fatigue Patterns at the Start, Middle, and End of Shifts by Hourly Categories by Date")
+if 'start' in df.columns and not df.empty:
+    try:
+        df_local = df.copy()
+        df_local['hour'] = df_local['start'].dt.hour
+        df_local['date'] = df_local['start'].dt.normalize()
+        # Kategorisasi jam menggunakan fungsi global
+        df_local['hour_category'] = df_local['hour'].apply(hour_range_label_full)
+        color_map = {
+            'Shift 1 Morning Early (6-9)':  '#FFEB3B',
+            'Shift 1 Morning Late (9-12)':   '#FFC107',
+            'Shift 1 Afternoon Early (12-15)':'#FF9800',
+            'Shift 1 Afternoon Late (15-18)': '#F57C00',
+            'Shift 2 Evening Early (18-21)':  '#42A5F5',
+            'Shift 2 Evening Late (21-24)':   '#1976D2',
+            'Shift 2 Dawn Early (0-3)':     '#0288D1',
+            'Shift 2 Dawn Late (3-6)':      '#01579B',
+            'Unknown': '#E0E0E0'
+        }
+        # Hitung jumlah fatigue per hari dan kategori jam
+        daily_by_cat = df_local.groupby(['date', 'hour_category']).size().reset_index(name='fatigue_count')
+        # --- TAMBAHAN: Ambil dominant hour_category per hari untuk Objective 3 ---
+        # Kita gunakan data df_local yang sudah memiliki hour_category
+        daily_dominant_cat = df_local.groupby('date')['hour_category'].agg(
+            lambda x: x.value_counts().idxmax()
+        ).reset_index()
+        daily_dominant_cat.rename(columns={'hour_category': 'dominant_hour_category'}, inplace=True)
+        # --- END TAMBAHAN ---
+        all_dates = pd.date_range(start=daily_by_cat['date'].min(), end=daily_by_cat['date'].max(), freq='D')
+        all_cats = list(color_map.keys())
+        full_index = pd.MultiIndex.from_product([all_dates, all_cats], names=['date', 'hour_category'])
+        daily_by_cat = daily_by_cat.set_index(['date', 'hour_category']).reindex(full_index, fill_value=0).reset_index()
+        daily_by_cat['day_of_week_num'] = daily_by_cat['date'].dt.dayofweek
+        daily_by_cat['week_start'] = daily_by_cat['date'] - pd.to_timedelta(daily_by_cat['day_of_week_num'], unit='d')
+        daily_by_cat['week_label'] = daily_by_cat['week_start'].dt.strftime('Week %U')
+        fig = px.bar(
+            daily_by_cat,
+            x='date',
+            y='fatigue_count',
+            color='hour_category',
+            title="Daily Fatigue Alerts by Detailed Hour Category",
+            color_discrete_map=color_map,
+            labels={'fatigue_count': 'Fatigue Alerts', 'date': 'Date'},
+            hover_data={'fatigue_count': True, 'week_label': True}
+        )
+        fig.update_layout(
+            barmode='stack',
+            xaxis_title="Date",
+            yaxis_title="Fatigue Alerts",
+            height=400,
+            legend_title="Hour Category"
+        )
+        unique_weeks = daily_by_cat['week_start'].unique()
+        shapes = []
+        week_labels = []
+        bg_colors = ['#f0e6ff', '#e6f0ff', '#e6fff0', '#fff0e6', '#ffe6e6', '#f0ffe6', '#e6e6ff']
+        for i, week in enumerate(sorted(unique_weeks)):
+            week_days = daily_by_cat[daily_by_cat['week_start'] == week]['date']
+            if len(week_days) > 0:
+                start_date = week_days.min()
+                end_date = week_days.max()
+                shapes.append(dict(
+                    type="rect",
+                    xref="x",
+                    yref="paper",
+                    x0=start_date,
+                    x1=end_date,
+                    y0=0,
+                    y1=1,
+                    fillcolor=bg_colors[i % len(bg_colors)],
+                    opacity=0.2,
+                    layer="below",
+                    line_width=0,
+                ))
+                week_labels.append(
+                    dict(
+                        xref='x',
+                        yref='paper',
+                        x=start_date + (end_date - start_date) / 2,
+                        y=1.02,
+                        text=f"Week {week.strftime('%U')}",
+                        showarrow=False,
+                        font=dict(size=10),
+                        xanchor='center',
+                        yanchor='bottom'
+                    )
+                )
+        fig.update_layout(shapes=shapes, annotations=week_labels)
+        st.plotly_chart(fig, use_container_width=True)
+    except Exception as e:
+        st.error(f"⚠️ Error in Daily Fatigue by Detailed Hour Category: {e}")
+else:
+    st.info("ℹ️ Insufficient time data to display this visualization.")
+
+# =================== OBJECTIVE 3: Daily Roster Insight per Week (Scatter Plot) =====================
+# =================== OBJECTIVE 3: Daily Roster Insight per Week (Scatter Plot) =====================
+st.subheader("OBJECTIVE 3: Daily Roster Insight per Week")
+if not df.empty and col_operator in df.columns and col_shift and col_shift in df.columns:
+    try:
+        df['date'] = pd.to_datetime(df['date'])
+        # Hitung total event per hari
+        daily_totals = df.groupby('date').size().reset_index(name='total_count')
+        # Ambil dominant shift per hari
+        dominant_shift = df.groupby('date')[col_shift].agg(lambda x: x.value_counts().idxmax()).reset_index()
+        dominant_shift.rename(columns={col_shift: 'dominant_shift'}, inplace=True)
+        daily_analysis = daily_totals.merge(dominant_shift, on='date', how='left')
+        daily_analysis['week_start'] = daily_analysis['date'] - pd.to_timedelta(daily_analysis['date'].dt.weekday, unit='d')
+        summary = []
+        weekly_groups = daily_analysis.groupby('week_start')
+        for week_start, week_data in weekly_groups:
+            # Urutkan data berdasarkan tanggal dalam minggu ini
+            week_data_sorted = week_data.sort_values('date').reset_index(drop=True)
+            for idx, row in week_data_sorted.iterrows():
+                current_date = row['date']
+                current_shift = row['dominant_shift']
+                current_count = row['total_count']
+                # --- CARI DATA DI HARI SEBELUM DAN SESUDAH (BERDASARKAN TANGGAL, BUKAN INDEKS) ---
+                prev_date = current_date - pd.Timedelta(days=1)
+                next_date = current_date + pd.Timedelta(days=1)
+                # Cari shift di hari sebelumnya
+                prev_row = week_data_sorted[week_data_sorted['date'] == prev_date]
+                prev_shift = prev_row['dominant_shift'].iloc[0] if not prev_row.empty else None
+                # Cari shift di hari berikutnya
+                next_row = week_data_sorted[week_data_sorted['date'] == next_date]
+                next_shift = next_row['dominant_shift'].iloc[0] if not next_row.empty else None
+                # ---- LOGIKA REMARK BERDASARKAN PERUBAHAN SHIFT DALAM MINGGU YANG SAMA----
+                # Awal Roster: Ada data di hari sebelumnya (prev_date) dalam minggu, dan shift-nya berbeda
+                # Akhir Roster: Ada data di hari berikutnya (next_date) dalam minggu, dan shift-nya berbeda
+                # Bukan Awal/Akhir: Ada data di hari sebelumnya ATAU berikutnya, dan shift-nya sama
+                # Unknown: Tidak ada data di hari sebelumnya (prev_date) DAN tidak ada data di hari berikutnya (next_date) dalam minggu yang sama
+                if pd.isna(current_shift):
+                    remark = "Unknown"
+                elif prev_shift is not None and prev_shift != current_shift:
+                    remark = "Start of Roster"
+                elif next_shift is not None and next_shift != current_shift:
+                    remark = "End of Roster"
+                elif (prev_shift is not None and prev_shift == current_shift) or (next_shift is not None and next_shift == current_shift):
+                    remark = "Neither Start nor End of Roster"
+                elif prev_shift is None and next_shift is None:
+                    remark = "Unknown"
+                else:
+                    remark = "Unknown"
+                # --- Operator dari data df (YANG SUDAH DIFILTER) ---
+                df_orig_for_date = df[df['date']==current_date] # Gunakan df yang difilter
+                if not df_orig_for_date.empty:
+                    peak_nik_counts = df_orig_for_date[col_operator].value_counts()
+                    peak_nik = peak_nik_counts.index[0] if not peak_nik_counts.empty else "N/A"
+                else:
+                    peak_nik = "N/A"
+                summary.append({
+                    'week_start': week_start,
+                    'date': current_date,
+                    'day_name': current_date.strftime('%A'),
+                    'total_count': current_count,
+                    'shift_category': current_shift,
+                    'remark': remark,
+                    'operator': peak_nik
+                })
+        summary_df = pd.DataFrame(summary)
+        if not summary_df.empty:
+            # Buat color map untuk remark (sesuai permintaan Anda)
+            color_map_remark = {
+                'Start of Roster': '#ffcccc',      # Merah muda
+                'End of Roster': '#cce5ff',     # Biru muda
+                'Neither Start nor End of Roster': '#fff2cc', # Kuning muda
+                'Unknown': '#c0c0c0'           # Abu-abu muda
+            }
+            # ===== SCATTER PLOT (WARNA BERDASARKAN remark) =====
+            fig = px.scatter(
+                summary_df,
+                x='date',
+                y='remark',
+                color='remark', # Warna berdasarkan remark (satu-satunya kolom di sumbu Y)
+                color_discrete_map=color_map_remark, # Gunakan color_map_remark
+                size='total_count',
+                hover_data=['shift_category', 'operator', 'total_count'],
+                title="Daily Roster Status by Date and Trend",
+                category_orders={'remark': ['Start of Roster', 'End of Roster', 'Neither Start nor End of Roster', 'Unknown']}
+            )
+            fig.update_layout(height=450, xaxis_title="Date", yaxis_title="Roster Status")
+            st.plotly_chart(fig, use_container_width=True)
+            # ===== TABEL =====
+            table_df = summary_df.rename(columns={
+                'week_start':'Week Start',
+                'day_name':'Day',
+                'date':'Date',
+                'total_count':'Event Count',
+                'shift_category':'Dominant Shift',
+                'remark':'Roster Status',
+                'operator':'Operator'
+            })
+            def highlight_remark(row):
+                colors = {
+                    'Start of Roster':'background-color: #ffcccc',
+                    'End of Roster':'background-color: #cce5ff',
+                    'Neither Start nor End of Roster':'background-color: #fff2cc',
+                    'Unknown':'background-color: #c0c0c0'
+                }
+                return [colors.get(row['Roster Status'], '') for _ in row]
+            st.dataframe(table_df.style.apply(highlight_remark, axis=1), use_container_width=True)
+        else:
+            st.info("ℹ️ No daily data to analyze.")
+    except Exception as e:
+        st.error(f"Error in Daily Roster Insight: {e}")
+else:
+    if col_shift is None:
+        st.info("ℹ️ Shift column not found, cannot display Daily Roster Insight.")
+    elif col_shift not in df.columns:
+        st.info(f"ℹ️ Column '{col_shift}' not found in the filtered data, cannot display Daily Roster Insight.")
+    else:
+        st.info("ℹ️ Insufficient data (date, operator, or shift column not found) to display daily roster insight.")
+# import plotly.express as px
+# from datetime import datetime
+st.subheader("OBJECTIVE 4: How is the Fatigue Event Risk Map per Operator?")
+import math
+import plotly.express as px
+try:
+    # ============================
+    # 1. PREPROCESS & COPY DF
+    # ============================
+    df_local = df.copy()
+    df_local['date_only'] = df_local['start'].dt.normalize()
+    df_local['week_number'] = df_local['date_only'].dt.isocalendar().week
+    df_local['week_label'] = "Week " + df_local['week_number'].astype(str)
+    # Unit cleanup
+    df_local['unit_no'] = (
+        df_local[col_fleet_no]
+        .astype(str)
+        .str.split("-", n=1).str[-1].str.strip()
+    )
+    if 'id' not in df_local.columns:
+        st.error("❌ Column 'id' not found!")
+        st.stop()
+    # ============================
+    # 2. FILTER 8 MINGGU TERAKHIR
+    # ============================
+    df_local['week_num_int'] = df_local['week_number'].astype(int)
+    unique_weeks = sorted(df_local['week_num_int'].unique())
+    selected_last8 = unique_weeks[-8:] if len(unique_weeks) >= 8 else unique_weeks
+    df_8w = df_local[df_local['week_num_int'].isin(selected_last8)].copy()
+    # =====================================
+    # 3. FREQUENCY PER OPERATOR PER MINGGU
+    # =====================================
+    weekly_freq = (
+        df_8w.groupby([col_operator, 'week_label'])['id']
+        .nunique()
+        .reset_index(name='weekly_frequency')
+    )
+    # ============================================
+    # 4. SUMMARY FREQUENCY & CEIL AVERAGE FREQ
+    # ============================================
+    freq_summary = (
+        weekly_freq
+        .groupby(col_operator)['weekly_frequency']
+        .agg(['sum', 'mean', 'count'])
+        .reset_index()
+        .rename(columns={
+            'sum': 'frequency_by_shift',
+            'mean': 'avg_frequency',
+            'count': 'frequency_by_weeks'
+        })
+    )
+    freq_summary['avg_frequency'] = freq_summary['avg_frequency'].apply(lambda x: math.ceil(x))
+    # ================================
+    # 5. RATA-RATA SPEED PER OPERATOR
+    # ================================
+    speed_summary = (
+        df_8w.groupby(col_operator)[col_speed]
+        .mean()
+        .reset_index(name='avg_speed')
+    )
+    # =====================
+    # 6. GABUNGKAN DATA
+    # =====================
+    risk_matrix = freq_summary.merge(speed_summary, on=col_operator, how='left')
+    risk_matrix = risk_matrix.rename(columns={col_operator: "Operator Name"})
+    # ================================
+    # 7. Tentukan Quadrant untuk Count
+    # ================================
+    def assign_quadrant(row):
+        if row['avg_frequency'] >= 2.5 and row['avg_speed'] >= 20:
+            return "Quadrant I – Prevent at Source"
+        elif row['avg_frequency'] < 2.5 and row['avg_speed'] >= 20:
+            return "Quadrant II – Detect & Monitor"
+        elif row['avg_frequency'] >= 2.5 and row['avg_speed'] < 20:
+            return "Quadrant III – Monitor"
+        else:
+            return "Quadrant IV – Low Control"
+    risk_matrix['quadrant'] = risk_matrix.apply(assign_quadrant, axis=1)
+    quadrant_count = risk_matrix['quadrant'].value_counts().reindex([
+        "Quadrant I – Prevent at Source",
+        "Quadrant II – Detect & Monitor",
+        "Quadrant III – Monitor",
+        "Quadrant IV – Low Control"
+    ], fill_value=0)
+    # ================================
+    # 8. VISUAL SCATTER PLOT
+    # ================================
+    fig = px.scatter(
+        risk_matrix,
+        x='avg_frequency',
+        y='avg_speed',
+        hover_name="Operator Name",
+        title="Operator Risk Matrix: Frequency vs Speed",
+        size=[12] * len(risk_matrix),
+        size_max=15
+    )
+    max_x = risk_matrix['avg_frequency'].max() + 1
+    max_y = risk_matrix['avg_speed'].max() + 1
+    # ================================
+    # 9. Quadrant Coloring
+    # ================================
+    fig.add_shape(type="rect", x0=2.5, x1=max_x, y0=20, y1=max_y,
+                  fillcolor="rgba(255,0,0,0.25)", line_width=0)   # I
+    fig.add_shape(type="rect", x0=0, x1=2.5, y0=20, y1=max_y,
+                  fillcolor="rgba(255,150,50,0.25)", line_width=0) # II
+    fig.add_shape(type="rect", x0=2.5, x1=max_x, y0=0, y1=20,
+                  fillcolor="rgba(255,200,200,0.25)", line_width=0) # III
+    fig.add_shape(type="rect", x0=0, x1=2.5, y0=0, y1=20,
+                  fillcolor="rgba(0,120,255,0.15)", line_width=0) # IV
+    # Garis batas
+    fig.add_vline(x=2.5, line_dash="dash", line_color="black")
+    fig.add_hline(y=20, line_dash="dash", line_color="black")
+    # ================================
+    # 10. Tampilkan Count di Quadrant
+    # ================================
+    fig.add_annotation(
+        x=2.5 + (max_x-2.5)/2, y=20 + (max_y-20)/2,
+        text=f"<b>{quadrant_count['Quadrant I – Prevent at Source']}</b>",
+        showarrow=False, font=dict(size=20, color="red")
+    )
+    fig.add_annotation(
+        x=2.5/2, y=20 + (max_y-20)/2,
+        text=f"<b>{quadrant_count['Quadrant II – Detect & Monitor']}</b>",
+        showarrow=False, font=dict(size=20, color="orange")
+    )
+    fig.add_annotation(
+        x=2.5 + (max_x-2.5)/2, y=0 + (20-0)/2,
+        text=f"<b>{quadrant_count['Quadrant III – Monitor']}</b>",
+        showarrow=False, font=dict(size=20, color="darkred")
+    )
+    fig.add_annotation(
+        x=2.5/2, y=0 + (20-0)/2,
+        text=f"<b>{quadrant_count['Quadrant IV – Low Control']}</b>",
+        showarrow=False, font=dict(size=20, color="blue")
+    )
+    # ================================
+    # 11. Label Quadrant
+    # ================================
+    fig.add_annotation(x=4, y=max_y-2, text="Quadrant I<br>Prevent at Source",
+                       showarrow=False, font=dict(size=12))
+    fig.add_annotation(x=1.25, y=max_y-2, text="Quadrant II<br>Detect & Monitor",
+                       showarrow=False, font=dict(size=12))
+    fig.add_annotation(x=4, y=5, text="Quadrant III<br>Monitor",
+                       showarrow=False, font=dict(size=12))
+    fig.add_annotation(x=1.25, y=5, text="Quadrant IV<br>Low Control",
+                       showarrow=False, font=dict(size=12))
+    fig.update_xaxes(dtick=1)
+    fig.update_layout(
+        xaxis_title="Average Frequency (Ceil)",
+        yaxis_title="Average Speed (km/h)",
+        height=650
+    )
+    st.plotly_chart(fig, use_container_width=True)
+    # ================================
+    # 12. DISPLAY TABLE
+    # ================================
+    st.subheader("Operator Risk Summary Table (8 Weeks Observed)")
+    table_display = (
+        risk_matrix[[
+            "Operator Name",
+            "frequency_by_shift",
+            "avg_frequency",
+            "frequency_by_weeks",
+            "avg_speed",
+            "quadrant"
+        ]]
+        .rename(columns={
+            "frequency_by_shift": "Frequency by Shift",
+            "avg_frequency": "Avg Frequency",
+            "frequency_by_weeks": "Frequency by Weeks",
+            "avg_speed": "Avg Speed"
+        })
+    )
+    st.dataframe(
+        table_display.sort_values("Avg Frequency", ascending=False),
+        use_container_width=True
+    )
+except Exception as e:
+    st.error(f"⚠️ Error Risk Map Objective 4: {e}")
+    st.exception(e)
+
+# ... (kode sebelumnya tetap sama) ...
+# ... (kode sebelumnya tetap sama) ...
+# ... (kode sebelumnya tetap sama) ...
+
+# =================== OBJECTIVE 5: Operator Fatigue Risk Gradient Dashboard =====================
+
+# ... (kode sebelumnya tetap sama) ...
+
+# =================== OBJECTIVE 5: Operator Fatigue Risk Gradient Dashboard =====================
+st.subheader("OBJECTIVE 5: Operator Fatigue Risk Gradient Dashboard (Weekly Average Events & Trend Analysis)")
+# Custom CSS untuk tampilan ala market saham yang sangat fancy dan profesional
+st.markdown("""
+<style>
+    .big-title {
+        font-size: 28px;
+        font-weight: bold;
+        color: #ffffff;
+        text-align: center;
+        margin-bottom: 10px;
+        background: linear-gradient(135deg, #2c3e50, #1a252c);
+        padding: 15px;
+        border-radius: 10px;
+        box-shadow: 0 4px 15px rgba(0,0,0,0.3);
+    }
+    .subnote {
+        font-size: 16px;
+        color: #7f8c8d;
+        text-align: center;
+        margin-bottom: 20px;
+    }
+    .section-divider {
+        height: 2px;
+        background: linear-gradient(to right, #3498db, #2ecc71, #f1c40f, #e74c3c);
+        margin: 20px 0;
+    }
+    .legend-container {
+        display: flex;
+        gap: 15px;
+        margin: 15px 0;
+    }
+    .legend-box {
+        background: white;
+        border: 1px solid #ddd;
+        border-radius: 8px;
+        padding: 15px;
+        flex: 1;
+        min-width: 300px;
+        box-shadow: 0 2px 10px rgba(0,0,0,0.05);
+    }
+    .legend-title {
+        font-weight: bold;
+        color: #2c3e50;
+        margin-bottom: 10px;
+        font-size: 14px;
+        border-bottom: 1px solid #eee;
+        padding-bottom: 5px;
+    }
+    .legend-item {
+        display: flex;
+        align-items: center;
+        margin: 5px 0;
+        font-size: 12px;
+    }
+    .legend-color {
+        width: 18px;
+        height: 18px;
+        border-radius: 3px;
+        margin-right: 8px;
+        border: 1px solid #ccc;
+    }
+    .ai-insight-box {
+        background: #f8f9fa;
+        border: 1px solid #dee2e6;
+        border-radius: 8px;
+        padding: 15px;
+        margin: 10px 0;
+        color: #2c3e50;
+        font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
+        box-shadow: 0 2px 8px rgba(0,0,0,0.05);
+    }
+    .ai-insight-title {
+        font-weight: bold;
+        color: #2c3e50;
+        margin-bottom: 8px;
+        font-size: 14px;
+        background: #e9ecef;
+        padding: 8px;
+        border-radius: 5px;
+        border-left: 4px solid #495057;
+    }
+    .trend-up {
+        color: #e74c3c;
+        font-weight: bold;
+    }
+    .trend-down {
+        color: #27ae60;
+        font-weight: bold;
+    }
+    .recommendation-box {
+        background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+        border: 1px solid #4a5568;
+        border-radius: 8px;
+        padding: 15px;
+        margin: 10px 0;
+        color: white;
+        font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
+        box-shadow: 0 4px 15px rgba(0,0,0,0.1);
+    }
+    .recommendation-title {
+        font-weight: bold;
+        color: white;
+        margin-bottom: 8px;
+        font-size: 14px;
+        background: rgba(255,255,255,0.2);
+        padding: 8px;
+        border-radius: 5px;
+        border-left: 4px solid white;
+    }
+    .recommendation-reason {
+        font-size: 12px;
+        margin-top: 10px;
+        padding: 8px;
+        background: rgba(255,255,255,0.1);
+        border-radius: 5px;
+        border-left: 3px solid rgba(255,255,255,0.3);
+    }
+</style>
+""", unsafe_allow_html=True)
+
+# ===============================================================
+# LOGIC UTAMA
+# ===============================================================
+if df.empty:
+    st.info("No data available after applying filters.")
+else:
+    try:
+        # Validasi kolom
+        required = [col_operator, col_fleet_type, "start"]
+        if not all(c in df.columns for c in required if c is not None):
+            st.warning("Required columns (operator, fleet_type, start) are missing.")
+            st.stop()
+
+        df_op = df[[col_operator, col_fleet_type, "start"]].dropna()
+        if df_op.empty:
+            st.info("No operator data after filtering.")
+            st.stop()
+
+        # Pastikan col_operator bukan None sebelum digunakan
+        if col_operator is None:
+            st.error(f"Operator column could not be auto-detected. Please check your data.")
+            st.stop()
+
+        df_op["year_week"] = df_op["start"].dt.strftime("%Y-W%U")
+
+        # Fuzzy match fleet names
+        fleet_clean = df_op[col_fleet_type].str.strip().str.upper()
+        df_op["is_ob"] = fleet_clean.str.contains(r"OB HAULLER", na=False)
+        df_op["is_coal"] = fleet_clean.str.contains(r"HAULING COAL", na=False)
+
+        ob_data = df_op[df_op["is_ob"]]
+        coal_data = df_op[df_op["is_coal"]]
+
+        # Fungsi hitung top 10 (untuk bar chart) - berdasarkan weekly avg events tertinggi
+        def get_top10_with_slope(data):
+            if data.empty:
+                st.warning("Data is empty in get_top10_with_slope.")
+                return pd.DataFrame()
+            # Pastikan col_operator tidak None dan ada di data
+            if col_operator is None or col_operator not in data.columns:
+                st.error(f"Operator column '{col_operator}' not found in data subset for get_top10.")
+                return pd.DataFrame()
+
+            weekly = data.groupby([col_operator, "year_week"]).size().reset_index(name="weekly_sum")
+            metrics = []
+            try:
+                for nik, grp in weekly.groupby(col_operator):
+                    # Lewati jika nik adalah None
+                    if pd.isna(nik):
+                        continue
+                    grp = grp.sort_values("year_week")
+                    counts = grp["weekly_sum"].values
+                    weeks = np.arange(len(counts))
+                    weekly_avg = counts.mean()
+                    total_events = counts.sum()
+                    n_weeks = len(counts)
+                    if n_weeks >= 2:
+                        x_mean = weeks.mean()
+                        y_mean = counts.mean()
+                        numerator = np.sum((weeks - x_mean) * (counts - y_mean))
+                        denominator = np.sum((weeks - x_mean) ** 2)
+                        slope = numerator / denominator if denominator != 0 else 0.0
+                    else:
+                        slope = 0.0
+                    metrics.append({
+                        col_operator: nik,
+                        "weekly_avg": weekly_avg,
+                        "slope": slope,
+                        "total_events": total_events,
+                        "n_weeks": n_weeks
+                    })
+            except KeyError as e:
+                st.error(f"KeyError in get_top10_with_slope: {e}. This might happen if the operator column contains invalid data types or unexpected values.")
+                return pd.DataFrame()
+            # Ambil top 10 berdasarkan weekly_avg (descending order)
+            if not metrics:
+                st.warning("No valid operator data found for slope calculation in get_top10.")
+                return pd.DataFrame()
+            return pd.DataFrame(metrics).nlargest(10, "weekly_avg")
+
+        top_ob = get_top10_with_slope(ob_data)
+        top_coal = get_top10_with_slope(coal_data)
+
+        # Fungsi hitung semua operator (untuk summary)
+        def get_all_operators_with_slope(data):
+            if data.empty:
+                st.warning("Data is empty in get_all_operators_with_slope.")
+                return pd.DataFrame()
+            # Pastikan col_operator tidak None dan ada di data
+            if col_operator is None or col_operator not in data.columns:
+                st.error(f"Operator column '{col_operator}' not found in data subset for get_all.")
+                return pd.DataFrame()
+
+            weekly = data.groupby([col_operator, "year_week"]).size().reset_index(name="weekly_sum")
+            metrics = []
+            try:
+                for nik, grp in weekly.groupby(col_operator):
+                    # Lewati jika nik adalah None
+                    if pd.isna(nik):
+                        continue
+                    grp = grp.sort_values("year_week")
+                    counts = grp["weekly_sum"].values
+                    weeks = np.arange(len(counts))
+                    weekly_avg = counts.mean()
+                    total_events = counts.sum()
+                    n_weeks = len(counts)
+                    if n_weeks >= 2:
+                        x_mean = weeks.mean()
+                        y_mean = counts.mean()
+                        numerator = np.sum((weeks - x_mean) * (counts - y_mean))
+                        denominator = np.sum((weeks - x_mean) ** 2)
+                        slope = numerator / denominator if denominator != 0 else 0.0
+                    else:
+                        slope = 0.0
+                    metrics.append({
+                        col_operator: nik,
+                        "weekly_avg": weekly_avg,
+                        "slope": slope,
+                        "total_events": total_events,
+                        "n_weeks": n_weeks
+                    })
+            except KeyError as e:
+                st.error(f"KeyError in get_all_operators_with_slope: {e}. This might happen if the operator column contains invalid data types or unexpected values.")
+                return pd.DataFrame()
+            if not metrics:
+                st.warning("No valid operator data found for slope calculation in get_all.")
+                return pd.DataFrame()
+            return pd.DataFrame(metrics)
+
+        all_ob = get_all_operators_with_slope(ob_data)
+        all_coal = get_all_operators_with_slope(coal_data)
+
+        # ===============================================================
+        # LEGEND DI LUAR CHART - 3 KOTAK DENGAN UKURAN SAMA
+        # ===============================================================
+        st.subheader("Risk Gradient Legend")
+        st.markdown("""
+        <div class="legend-container">
+            <div class="legend-box">
+                <div class="legend-title">Worsening Trends (Positive Slope):</div>
+                <div class="legend-item">
+                    <div class="legend-color" style="background-color: #d32f2f;"></div>
+                    <span>Very High Risk (≥1.5)</span>
+                </div>
+                <div class="legend-item">
+                    <div class="legend-color" style="background-color: #e57373;"></div>
+                    <span>High Risk (1.0-1.5)</span>
+                </div>
+                <div class="legend-item">
+                    <div class="legend-color" style="background-color: #ef9a9a;"></div>
+                    <span>Moderate Risk (0.5-1.0)</span>
+                </div>
+                <div class="legend-item">
+                    <div class="legend-color" style="background-color: #ffcdd2;"></div>
+                    <span>Slight Risk (0-0.5)</span>
+                </div>
+            </div>
+            <div class="legend-box">
+                <div class="legend-title">Improving Trends (Negative Slope):</div>
+                <div class="legend-item">
+                    <div class="legend-color" style="background-color: #388e3c;"></div>
+                    <span>Excellent Improvement (≤-1.5)</span>
+                </div>
+                <div class="legend-item">
+                    <div class="legend-color" style="background-color: #81c784;"></div>
+                    <span>Great Improvement (-1.5 to -1.0)</span>
+                </div>
+                <div class="legend-item">
+                    <div class="legend-color" style="background-color: #a5d6a7;"></div>
+                    <span>Good Improvement (-1.0 to -0.5)</span>
+                </div>
+                <div class="legend-item">
+                    <div class="legend-color" style="background-color: #c8e6c9;"></div>
+                    <span>Slight Improvement (-0.5-0)</span>
+                </div>
+            </div>
+            <div class="legend-box">
+                <div class="legend-title">Stable Trend (Zero Slope):</div>
+                <div class="legend-item">
+                    <div class="legend-color" style="background-color: #95a5a6;"></div>
+                    <span>Stable (0)</span>
+                </div>
+                <br>
+                <i>Note: Only appears when operator data shows consistent behavior within a single week observation period.</i>
+            </div>
+        </div>
+        """, unsafe_allow_html=True)
+
+        # ===============================================================
+        # PLOT FUNCTION (Bar Chart with Risk Gradient Colors) - PERBAIKAN DI SINI
+        # ===============================================================
+        def plot_chart(data, title):
+            if data.empty:
+                fig = go.Figure()
+                fig.add_annotation(
+                    text="No Data",
+                    x=0.5, y=0.5,
+                    showarrow=False,
+                    font_size=16
+                )
+                # Gunakan update_layout untuk menetapkan judul
+                fig.update_layout(height=350, title=title)
+                return fig
+
+            # Urutkan data berdasarkan weekly_avg dari besar ke kecil
+            data_sorted = data.sort_values('weekly_avg', ascending=False)
+
+            # Kategorisasi warna berdasarkan slope dengan gradasi yang berbeda
+            def get_color(slope):
+                if slope == 0:
+                    return "#95a5a6"  # Abu-abu (Stabil)
+                elif slope > 0:
+                    # Gradasi merah untuk slope positif
+                    if slope < 0.5:
+                        return "#ffcdd2"  # Merah sangat muda
+                    elif slope < 1.0:
+                        return "#ef9a9a"  # Merah muda
+                    elif slope < 1.5:
+                        return "#e57373"  # Merah sedang
+                    else:
+                        return "#d32f2f"  # Merah gelap
+                else:  # slope < 0
+                    # Gradasi hijau untuk slope negatif
+                    if slope > -0.5:
+                        return "#c8e6c9"  # Hijau sangat muda
+                    elif slope > -1.0:
+                        return "#a5d6a7"  # Hijau muda
+                    elif slope > -1.5:
+                        return "#81c784"  # Hijau sedang
+                    else:
+                        return "#388e3c"  # Hijau gelap
+
+            colors = [get_color(s) for s in data_sorted["slope"]]
+
+            # Buat trace bar, TANPA argumen 'title'
+            bar_trace = go.Bar(
+                x=data_sorted[col_operator].astype(str),
+                y=data_sorted["weekly_avg"],
+                marker=dict(
+                    color=colors,
+                    line=dict(width=2, color="rgba(0,0,0,0.2)")
+                ),
+                text=[f"{v:.1f}" for v in data_sorted["weekly_avg"]],
+                textposition="outside",
+                hovertemplate=(
+                    "<b>%{x}</b><br>" +
+                    "Weekly Avg: %{y:.2f}<br>" +
+                    "Trend Slope: %{customdata[0]:+.3f}<br>" +
+                    "Total Events: %{customdata[1]}<br>" +
+                    "Weeks Active: %{customdata[2]}<br>" +
+                    "<extra></extra>"
+                ),
+                customdata=np.stack([data_sorted["slope"], data_sorted["total_events"], data_sorted["n_weeks"]], axis=-1)
+            )
+
+            # Buat figure dan tambahkan trace
+            fig = go.Figure(bar_trace)
+
+            # Gunakan update_layout untuk menetapkan judul dan layout lainnya
+            fig.update_layout(
+                title=f"<b>{title}</b>",
+                title_x=0.5, # Pusatkan judul
+                height=450,
+                margin=dict(l=50, r=20, t=60, b=120),
+                xaxis_title="<b>Operator ID</b>",
+                yaxis_title="<b>Weekly Avg Events</b>",
+                font=dict(family="Segoe UI", size=12),
+                bargap=0.3,
+                plot_bgcolor="rgba(0,0,0,0)",
+                paper_bgcolor="rgba(0,0,0,0)"
+            )
+            return fig
+
+        # ===============================================================
+        # TAMPILKAN BAR CHART
+        # ===============================================================
+        col1, col2 = st.columns(2)
+        with col1:
+            st.plotly_chart(plot_chart(top_ob, "OB HAULER Operators (Risk Gradient)"), use_container_width=True)
+        with col2:
+            st.plotly_chart(plot_chart(top_coal, "HAULING COAL Operators (Risk Gradient)"), use_container_width=True)
+
+        # ===============================================================
+        # AI INSIGHTS - DIBEDAKAN UNTUK OB HAULER DAN COAL HAULING - SEKARANG BERSEBELAHAN
+        # ===============================================================
+        st.markdown("---")
+        st.subheader("Data Insight Automation")
+
+        # Gunakan kolom untuk menampilkan analisis secara bersebelahan
+        col_insight1, col_insight2 = st.columns(2)
+
+        # Insight untuk OB HAULER - Ditampilkan di kolom kiri
+        with col_insight1:
+            if not top_ob.empty:
+                st.markdown("### OB HAULER Analysis")
+                ob_worsening = len(top_ob[top_ob['slope'] > 0])
+                ob_improving = len(top_ob[top_ob['slope'] < 0])
+                ob_avg_risk = top_ob['weekly_avg'].mean()
+                ob_max_risk = top_ob['weekly_avg'].max()
+                ob_insights = []
+                if ob_worsening > ob_improving:
+                    ob_insights.append(f"{ob_worsening} out of 10 top risk operators are showing <span class='trend-up'>worsening</span> trends, indicating potential fatigue issues in this fleet type.")
+                else:
+                    ob_insights.append(f"{ob_improving} out of 10 top risk operators are showing <span class='trend-down'>improvement</span>, suggesting effective fatigue management strategies.")
+                ob_insights.append(f"Average risk level among top 10 operators is {ob_avg_risk:.2f} events per week with maximum {ob_max_risk:.2f}.")
+
+                for insight in ob_insights:
+                    st.markdown(f"""
+                    <div class="ai-insight-box">
+                        <div class="ai-insight-title">Risk Analysis</div>
+                        <p>{insight}</p>
+                    </div>
+                    """, unsafe_allow_html=True)
+            else:
+                st.info("No OB HAULER data for analysis.")
+
+        # Insight untuk HAULING COAL - Ditampilkan di kolom kanan
+        with col_insight2:
+            if not top_coal.empty:
+                st.markdown("### HAULING COAL Analysis")
+                coal_worsening = len(top_coal[top_coal['slope'] > 0])
+                coal_improving = len(top_coal[top_coal['slope'] < 0])
+                coal_avg_risk = top_coal['weekly_avg'].mean()
+                coal_max_risk = top_coal['weekly_avg'].max()
+                coal_insights = []
+                if coal_worsening > coal_improving:
+                    coal_insights.append(f"{coal_worsening} out of 10 top risk operators are showing <span class='trend-up'>worsening</span> trends, requiring immediate attention.")
+                else:
+                    coal_insights.append(f"{coal_improving} out of 10 top risk operators are showing <span class='trend-down'>improvement</span>, indicating positive trends in safety management.")
+                coal_insights.append(f"Average risk level among top 10 operators is {coal_avg_risk:.2f} events per week with maximum {coal_max_risk:.2f}.")
+
+                for insight in coal_insights:
+                    st.markdown(f"""
+                    <div class="ai-insight-box">
+                        <div class="ai-insight-title">Risk Analysis</div>
+                        <p>{insight}</p>
+                    </div>
+                    """, unsafe_allow_html=True)
+            else:
+                st.info("No HAULING COAL data for analysis.")
+
+        # ===============================================================
+        # AI RECOMMENDATIONS - JUGA BERSEBELAHAN
+        # ===============================================================
+        st.markdown("---")
+        st.subheader("AI Recommendations")
+
+        # Gunakan kolom untuk menampilkan rekomendasi secara bersebelahan
+        col_rec1, col_rec2 = st.columns(2)
+
+        def generate_recommendations(top_ob, top_coal):
+            recommendations = {}
+            if not top_ob.empty:
+                ob_worsening = len(top_ob[top_ob['slope'] > 0])
+                ob_avg_risk = top_ob['weekly_avg'].mean()
+                if ob_worsening > 5:  # Lebih dari setengah
+                    recommendations['ob'] = "Implement immediate fatigue monitoring protocols for operators showing worsening trends."
+                    reason_ob = "High percentage of operators showing increasing risk trends indicates potential systemic fatigue issues requiring immediate intervention."
+                elif ob_avg_risk > 10:  # High average risk
+                    recommendations['ob'] = "Consider workload redistribution to reduce average risk levels."
+                    reason_ob = "High average risk levels suggest operational adjustments are needed to maintain optimal safety standards."
+                else:
+                    recommendations['ob'] = "Continue current safety protocols with enhanced monitoring for early detection."
+                    reason_ob = "Stable risk profile indicates current protocols are effective, but continuous monitoring ensures early detection of potential issues."
+                recommendations['ob_reason'] = reason_ob
+
+            if not top_coal.empty:
+                coal_worsening = len(top_coal[top_coal['slope'] > 0])
+                coal_avg_risk = top_coal['weekly_avg'].mean()
+                if coal_worsening > 5:  # Lebih dari setengah
+                    recommendations['coal'] = "Implement immediate fatigue monitoring protocols for operators showing worsening trends."
+                    reason_coal = "High percentage of operators showing increasing risk trends indicates potential systemic fatigue issues requiring immediate intervention."
+                elif coal_avg_risk > 10:  # High average risk
+                    recommendations['coal'] = "Consider workload redistribution to reduce average risk levels."
+                    reason_coal = "High average risk levels suggest operational adjustments are needed to maintain optimal safety standards."
+                else:
+                    recommendations['coal'] = "Continue current safety protocols with enhanced monitoring for early detection."
+                    reason_coal = "Stable risk profile indicates current protocols are effective, but continuous monitoring ensures early detection of potential issues."
+                recommendations['coal_reason'] = reason_coal
+
+            return recommendations
+
+        ai_recommendations = generate_recommendations(top_ob, top_coal)
+
+        # Recommendation untuk OB HAULER - Ditampilkan di kolom kiri
+        with col_rec1:
+            if 'ob' in ai_recommendations:
+                st.markdown("### OB HAULER Recommendations")
+                st.markdown(f"""
+                <div style="background: linear-gradient(135deg, #667eea 0%, #764ba2 100%); border: 1px solid #4a5568; border-radius: 8px; padding: 15px; margin: 10px 0; color: white; font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif; box-shadow: 0 4px 15px rgba(0,0,0,0.1); height: 150px; display: flex; flex-direction: column; justify-content: space-between;">
+                    <div style="font-weight: bold; font-size: 14px; background: rgba(255,255,255,0.2); padding: 8px; border-radius: 5px; border-left: 4px solid white;">Recommendation</div>
+                    <div style="padding-top: 8px; font-size: 14px;">{ai_recommendations['ob']}</div>
+                    <div style="font-size: 12px; margin-top: 10px; padding: 8px; background: rgba(255,255,255,0.1); border-radius: 5px; border-left: 3px solid rgba(255,255,255,0.3);">AI Reasoning: {ai_recommendations['ob_reason']}</div>
+                </div>
+                """, unsafe_allow_html=True)
+            else:
+                st.info("No OB HAULER recommendations generated.")
+
+        # Recommendation untuk HAULING COAL - Ditampilkan di kolom kanan
+        with col_rec2:
+            if 'coal' in ai_recommendations:
+                st.markdown("### HAULING COAL Recommendations")
+                st.markdown(f"""
+                <div style="background: linear-gradient(135deg, #667eea 0%, #764ba2 100%); border: 1px solid #4a5568; border-radius: 8px; padding: 15px; margin: 10px 0; color: white; font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif; box-shadow: 0 4px 15px rgba(0,0,0,0.1); height: 150px; display: flex; flex-direction: column; justify-content: space-between;">
+                    <div style="font-weight: bold; font-size: 14px; background: rgba(255,255,255,0.2); padding: 8px; border-radius: 5px; border-left: 4px solid white;">Recommendation</div>
+                    <div style="padding-top: 8px; font-size: 14px;">{ai_recommendations['coal']}</div>
+                    <div style="font-size: 12px; margin-top: 10px; padding: 8px; background: rgba(255,255,255,0.1); border-radius: 5px; border-left: 3px solid rgba(255,255,255,0.3);">AI Reasoning: {ai_recommendations['coal_reason']}</div>
+                </div>
+                """, unsafe_allow_html=True)
+            else:
+                st.info("No HAULING COAL recommendations generated.")
+
+    except Exception as e:
+        st.error(f"Error in Top 10 Operator analysis: {str(e)}")
+        st.code(f"Error: {e}", language="python")
+
+# ... (kode setelah Objective 5 tetap sama) ...
+
+# =================== OBJECTIVE 6: Automated Insights & AI Recommendations =====================
+st.subheader("OBJECTIVE 6: Automated Insight Summary & AI Recommendations")
+
+# Membagi tampilan menjadi dua kolom
+col_insights, col_recs = st.columns(2)
+
+# Kolom kiri: Insights by Advanced Analytics
+with col_insights:
+    st.subheader("Insights by Advanced Analytics")
+
+    # 1. Critical Hour Analysis (2-5 AM)
+    critical_hours = [2, 3, 4, 5]
+    critical_alerts = df[df['hour'].isin(critical_hours)]
+    critical_pct = (len(critical_alerts) / len(df)) * 100 if len(df) > 0 else 0
+
+    st.markdown(f"**Critical Hour Risk (2-5 AM)**")
+    # Use conditional formatting for background color
+    bg_color = "#ffcccc" if critical_pct > 50 else "#ffebcc" if critical_pct > 25 else "#ffffcc" if critical_pct > 10 else "#e6ffe6"
+    st.markdown(f'<div style="background-color: {bg_color}; padding: 10px; border-radius: 5px;">Critical Hour Alerts: {len(critical_alerts)} ({critical_pct:.1f}% of total alerts)</div>', unsafe_allow_html=True)
+    if critical_pct > 10:  # If more than 10% of alerts happen in critical hours
+        st.warning(f"High risk: {critical_pct:.1f}% of fatigue alerts occur during critical hours (2-5 AM). This is a known circadian dip period.")
+    else:
+        st.info(f"{critical_pct:.1f}% of alerts occur during critical hours. This is within acceptable range.")
+
+    # 2. High-Speed Fatigue Analysis (Environmental Risk)
+    if col_speed and col_speed in df.columns:
+        high_speed_threshold = df[col_speed].quantile(0.75) if not df[col_speed].dropna().empty else 0  # Handle empty series
+        high_speed_fatigue = df[df[col_speed] >= high_speed_threshold] if high_speed_threshold > 0 else pd.DataFrame()
+        high_speed_pct = (len(high_speed_fatigue) / len(df)) * 100 if len(df) > 0 else 0
+        
+        st.markdown(f"**High-Speed Fatigue Risk (Speed > {high_speed_threshold:.0f} km/h)**")
+        st.metric("High-Speed Fatigue Events", f"{len(high_speed_fatigue)}", f"{high_speed_pct:.1f}% of total alerts")
+        if high_speed_pct > 20:  # If more than 20% of alerts happen at high speed
+            st.warning(f"High risk: {high_speed_pct:.1f}% of fatigue alerts occur at high speeds. This increases accident severity potential.")
+        else:
+            st.info(f"{high_speed_pct:.1f}% of alerts occur at high speeds. This is within acceptable range.")
+    else:
+        st.info("Speed data not available for High-Speed Fatigue Analysis.")
+
+    # 3. Shift Pattern Analysis
+    if col_shift and col_shift in df.columns:
+        shift_counts = df[col_shift].value_counts()
+        # shift_alerts_by_hour = df.groupby([col_shift, 'hour']).size().reset_index(name='alerts') # Tidak digunakan dalam tampilan ini
+        
+        st.markdown(f"**Shift Pattern Risk**")
+        for shift_val in shift_counts.index:
+            shift_pct = (shift_counts[shift_val] / len(df)) * 100
+            st.metric(f"Shift {shift_val} Alerts", f"{shift_counts[shift_val]}", f"{shift_pct:.1f}% of total alerts")
+            if shift_pct > 50:  # If one shift has more than 50% of alerts
+                st.warning(f"Shift {shift_val} has disproportionately high alerts ({shift_pct:.1f}%). Review shift scheduling and workload.")
+            else:
+                st.info(f"Shift {shift_val} alert distribution is acceptable ({shift_pct:.1f}%).")
+    else:
+        st.info("Shift data not available for Shift Pattern Analysis.")
+
+    # 4. Operator Risk Profiling
+    if col_operator and col_operator in df.columns:
+        operator_alerts = df[col_operator].value_counts()
+        top_risk_operators = operator_alerts.head(5)  # Top 5 operators by alerts
+        
+        st.markdown(f"**High-Risk Operator Identification**")
+        for op_name, count in top_risk_operators.items():
+            op_pct = (count / len(df)) * 100
+            st.metric(f"Operator: {op_name}", f"{count} alerts", f"{op_pct:.1f}% of total alerts")
+            if op_pct > 5:  # If an operator has more than 5% of all alerts
+                st.warning(f"Operator {op_name} has high fatigue risk ({op_pct:.1f}% of alerts). Consider coaching or rest plan.")
+            else:
+                st.info(f"Operator {op_name} fatigue risk is within acceptable range ({op_pct:.1f}%).")
+    else:
+        st.info("Operator data not available for Operator Risk Profiling.")
+
+
+# Kolom kanan: AI Recommendations
+with col_recs:
+    st.subheader("AI Recommendations")
+    ai_recs = []
+    insights_found = [] # Untuk menyimpan insight yang ditemukan
+
+    # Peak hour
+    if "hour" in df.columns and not df.empty:
+        peak_hour = df["hour"].value_counts().idxmax()
+        critical_hours = [2, 3, 4, 5]
+        if peak_hour in critical_hours:
+            insights_found.append(f" Most fatigue risk occurs at **{peak_hour}:00** — during critical circadian low period (2-5 AM). Consider enhanced monitoring.")
+        else:
+            insights_found.append(f"Most fatigue risk occurs at **{peak_hour}:00** — likely due to circadian drop.")
+
+    # Risk shift
+    if col_shift and not df.empty:
+        worst_shift = df[col_shift].value_counts().idxmax()
+        insights_found.append(f" Highest fatigue recorded in **Shift {worst_shift}** — review scheduling & workload.")
+
+    # Worst operator
+    if col_operator and not df.empty:
+        worst_operator = df[col_operator].value_counts().idxmax()
+        insights_found.append(f" Operator at highest risk: **{worst_operator}** — suggested coaching or rest plan.")
+
+    # Duration risk
+    if "duration_sec" in df.columns and not df.empty:
+        avg_duration = df["duration_sec"].mean()
+        if not pd.isna(avg_duration) and avg_duration > 10:
+            insights_found.append(" Long fatigue event duration suggests slow response — improve alerting training.")
+
+    # Generate recommendations based on found insights
+    if insights_found:
+        # Contoh rekomendasi berdasarkan insight
+        if any("circadian low" in i.lower() for i in insights_found):
+            ai_recs.append({
+                "recommendation": "Deploy enhanced fatigue monitoring systems (e.g., EOR) specifically during 2-5 AM shifts.",
+                "data_point": f"Critical Hour Alerts: {len(critical_alerts)} ({critical_pct:.1f}% of total alerts)",
+                "reason": "High percentage of alerts occurring during the known circadian low period (2-5 AM) indicates increased risk during these hours."
+            })
+        if any("shift" in i.lower() for i in insights_found):
+            ai_recs.append({
+                "recommendation": "Review shift rotation schedules to minimize consecutive high-risk shifts.",
+                "data_point": f"Shift {worst_shift} Alerts: {df[col_shift].value_counts()[worst_shift]} ({(df[col_shift].value_counts()[worst_shift] / len(df)) * 100:.1f}% of total alerts)",
+                "reason": f"The identified high-risk shift ({worst_shift}) has the highest number of fatigue alerts, suggesting scheduling or workload issues."
+            })
+        if any("operator" in i.lower() for i in insights_found):
+            ai_recs.append({
+                "recommendation": "Initiate individual coaching or mandatory rest periods for high-risk operators.",
+                "data_point": f"Operator {worst_operator} Alerts: {df[col_operator].value_counts()[worst_operator]} ({(df[col_operator].value_counts()[worst_operator] / len(df)) * 100:.1f}% of total alerts)",
+                "reason": f"The identified high-risk operator ({worst_operator}) has the highest number of fatigue alerts, indicating a need for targeted intervention."
+            })
+        if any("duration" in i.lower() for i in insights_found):
+            ai_recs.append({
+                "recommendation": "Review and improve alert response protocols and training.",
+                "data_point": f"Average Fatigue Event Duration: {avg_duration:.2f} seconds",
+                "reason": "Long average duration suggests potential delays in response time or alert acknowledgment, requiring protocol review."
+            })
+        if any("high-speed" in i.lower() for i in insights_found):
+            ai_recs.append({
+                "recommendation": "Implement speed management strategies in conjunction with fatigue monitoring.",
+                "data_point": f"High-Speed Fatigue Events: {len(high_speed_fatigue)} ({high_speed_pct:.1f}% of total alerts)",
+                "reason": "A significant percentage of alerts occur at high speeds, increasing accident severity risk. Speed control is crucial."
+            })
+        if not ai_recs:
+            ai_recs.append({
+                "recommendation": "Data quality is sufficient. Focus on implementing recommendations from Objectives 1-5.",
+                "data_point": "General Data Quality Check",
+                "reason": "No specific high-impact insights were automatically identified from the aggregated data in this section."
+            })
+
+        # Menampilkan rekomendasi dalam format kotak yang sesuai dengan permintaan
+        for rec in ai_recs:
+            # Gunakan div dengan class khusus untuk membuat kotak rekomendasi di kolom kanan
+            # Gaya diambil dari .insight-box untuk konsistensi dan menghindari warna ungu
+            st.markdown(f"""
+            <div style="
+                background: #f8f9fa;
+                border: 1px solid #dee2e6;
+                border-radius: 8px;
+                padding: 15px;
+                margin: 10px 0;
+                color: #2c3e50;
+                font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
+                box-shadow: 0 2px 8px rgba(0,0,0,0.05);
+                display: flex;
+                flex-direction: column;
+                justify-content: space-between;
+            ">
+                <div style="
+                    font-weight: bold;
+                    color: #2c3e50;
+                    margin-bottom: 8px;
+                    font-size: 14px;
+                    background: #e9ecef;
+                    padding: 8px;
+                    border-radius: 5px;
+                    border-left: 4px solid #495057;
+                ">AI Recommendation</div>
+                <div style="padding-top: 8px; font-size: 14px; margin-bottom: 10px;">
+                    <strong>Action:</strong> {rec['recommendation']}
+                </div>
+                <div style="font-size: 12px; padding: 8px; background: #e9ecef; border-radius: 5px; margin-top: 5px;">
+                    <strong>Data Point:</strong> {rec['data_point']}
+                </div>
+                <div style="font-size: 12px; padding: 8px; background: #f1f1f1; border-radius: 5px; margin-top: 5px;">
+                    <strong>AI Reasoning:</strong> {rec['reason']}
+                </div>
+            </div>
+            """, unsafe_allow_html=True)
+    else:
+        st.info("No specific data points available for AI recommendations. Ensure relevant columns (hour, shift, operator, duration, speed) are present and populated.")
+
+# ================= FOOTER ===========================
+st.markdown("---")
+st.markdown('<div class="footer">MineVision AI - Transforming Mining Safety with Intelligent Analytics | Contact: info@bukittechnology.com</div>', unsafe_allow_html=True)
+
+
+# ================= FOOTER ===========================
+st.markdown("---")
+st.markdown('<div class="footer">MineVision AI - Transforming Mining Safety with Intelligent Analytics | Contact: info@bukittechnology.com</div>', unsafe_allow_html=True)

requirements.txt

@@ -1,3 +1,8 @@
 altair
-pandas
-streamlit
+streamlit>=1.38.0
+pandas>=2.2.2
+numpy>=1.26.4
+plotly>=5.24.1
+plotly-express>=0.4.1
+openpyxl>=3.1.5
+python-dateutil>=2.9.0