import streamlit as st
import joblib
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from prediction import predict
import os

# ================================
# Load model & preprocessor
# ================================
BASE_DIR = os.path.dirname(os.path.abspath(__file__))

model = joblib.load(os.path.join(BASE_DIR, 'rf_model.pkl'))
prep = joblib.load(os.path.join(BASE_DIR, 'preprocessor.pkl'))

# ================================
# Sidebar Navigation
# ================================
st.sidebar.title("Navigation")
page = st.sidebar.selectbox("Choose Page", ["Prediction", "EDA"])

st.sidebar.markdown("### Created by Fernando Brian")
st.sidebar.markdown("### Deployed on Hugging Face")

# ================================
# 📊 PAGE 1: PREDICTION
# ================================
if page == "Prediction":

    st.title("Hotel Booking Cancellation Prediction")

    st.write("Masukkan data booking untuk memprediksi kemungkinan pembatalan.")

    # ================================
    # Input User
    # ================================
    lead_time = st.slider("Lead Time", 0, 365, 50)
    hotel = st.selectbox("Hotel Type", ["City Hotel", "Resort Hotel"])
    deposit_type = st.selectbox("Deposit Type", ["No Deposit", "Non Refund", "Refundable"])
    market_segment = st.selectbox("Market Segment", ["Online TA", "Offline TA/TO", "Direct", "Corporate"])
    country = st.text_input("Country (contoh: PRT, GBR)", "PRT")

    # ================================
    # Prediction Button
    # ================================
    if st.button("Predict"):

        data = {
            'hotel': hotel,
            'lead_time': lead_time,
            'arrival_date_year': 2017,
            'arrival_date_month': 'July',
            'arrival_date_week_number': 27,
            'arrival_date_day_of_month': 1,
            'stays_in_weekend_nights': 1,
            'stays_in_week_nights': 2,
            'adults': 2,
            'children': 0,
            'babies': 0,
            'meal': 'BB',
            'country': country,
            'market_segment': market_segment,
            'distribution_channel': 'TA/TO',
            'is_repeated_guest': 0,
            'previous_cancellations': 0,
            'previous_bookings_not_canceled': 0,
            'reserved_room_type': 'A',
            'assigned_room_type': 'A',
            'booking_changes': 0,
            'deposit_type': deposit_type,
            'days_in_waiting_list': 0,
            'customer_type': 'Transient',
            'adr': 100.0,
            'required_car_parking_spaces': 0,
            'total_of_special_requests': 1,
            'agent': 0,
            'company': 0
        }

        # Predict
        result = predict(data, model, prep)

        # Output
        if result == 1:
            st.error("❌ Booking kemungkinan akan dibatalkan")
        else:
            st.success("✅ Booking kemungkinan tidak dibatalkan")

# ================================
# 📈 PAGE 2: EDA
# ================================
elif page == "EDA":

    st.title("Exploratory Data Analysis")

    # Load dataset
    df = pd.read_csv(os.path.join(BASE_DIR, "hotel_bookings.csv"))

    # ================================
    # Dataset Preview
    # ================================
    st.subheader("Dataset Preview")
    st.dataframe(df.head())

    # ================================
    # Plot 1 - Cancellation Distribution
    # ================================
    st.subheader("Distribusi Pembatalan")

    fig, ax = plt.subplots()
    sns.countplot(x='is_canceled', data=df, ax=ax)
    ax.set_title("Cancellation Distribution")
    st.pyplot(fig)

    # ================================
    # Plot 2 - Lead Time vs Cancellation
    # ================================
    st.subheader("Lead Time vs Cancellation")

    fig, ax = plt.subplots()
    sns.boxplot(x='is_canceled', y='lead_time', data=df, ax=ax)
    ax.set_title("Lead Time vs Cancellation")
    st.pyplot(fig)

    # ================================
    # Plot 3 - Market Segment Distribution
    # ================================
    st.subheader("Market Segment Distribution")

    fig, ax = plt.subplots()
    sns.countplot(
        y='market_segment',
        data=df,
        order=df['market_segment'].value_counts().index,
        ax=ax
    )
    ax.set_title("Market Segment Distribution")
    st.pyplot(fig)