Update app.py

app.py

@@ -723,90 +723,100 @@ def generate_objective2_insights(df: pd.DataFrame) -> str:
 
 # === RENDER INSIGHT BOX ===
 # =============== INSIGHT 2 (Ringkas & Fokus ke Red & Amber) ===============
-if alarm_data.empty:
-    insight_lines = ["• No alarm data available for Red or Amber analysis."]
+# =============== INSIGHT 3 (Objective 3: Red & Amber Alarm Patterns by Time & Position) ===============
+
+# Pastikan alarm_data ada dan punya kolom wajib
+required_cols = {'Alarm Status', 'hour', 'Position'}
+if not required_cols.issubset(alarm_data.columns) or alarm_data.empty:
+    insight_lines = [
+        "• No alarm data available for Red or Amber analysis.",
+        "• Ensure dataset includes 'Alarm Status', 'hour', and 'Position' columns."
+    ]
 else:
-    # Filter hanya Red dan Amber (case-insensitive, robust terhadap NaN)
-    red_amber_data = alarm_data[
-        alarm_data['Alarm Status'].str.contains(r'\b(Red|Amber)\b', case=False, na=False)
-    ].copy()
+    # Filter Red & Amber — case-insensitive, aman terhadap NaN
+    mask_red = alarm_data['Alarm Status'].str.contains(r'\bRed\b', case=False, na=False)
+    mask_amber = alarm_data['Alarm Status'].str.contains(r'\bAmber\b', case=False, na=False)
+    red_amber_data = alarm_data[mask_red | mask_amber].copy()
 
     if red_amber_data.empty:
         insight_lines = ["• No Red or Amber alarms detected in the dataset."]
     else:
-        # Hitung jumlah
-        red_alarms = red_amber_data['Alarm Status'].str.contains('Red', case=False).sum()
-        amber_alarms = red_amber_data['Alarm Status'].str.contains('Amber', case=False).sum()
-        total_red_amber = len(red_amber_data)
-
-        # Kelompok waktu dominan: hanya 2 band utama (sesuai preferensi)
-        def hour_to_band(h):
-            if 12 <= h < 18:
-                return "12:00–18:00 (Afternoon)"
-            elif (18 <= h <= 23) or (0 <= h < 6):
-                return "18:00–00:00 (Evening/Night)"
-            else:
-                return "06:00–12:00 (Morning)"  # fallback, jarang muncul
+        insight_lines = []
 
-        red_amber_data['time_band'] = red_amber_data['hour'].apply(hour_to_band)
-        band_counts = red_amber_data['time_band'].value_counts()
-        top2 = band_counts.nlargest(2)
+        # Helper: count alarms per (position, hour, type)
+        # — tanpa def, langsung vectorized
+        red_amber_data['is_red'] = mask_red.loc[red_amber_data.index]
+        red_amber_data['is_amber'] = mask_amber.loc[red_amber_data.index]
 
-        # Ambil dominan & kedua dominan (jika ada)
-        dominant_band = top2.index[0] if len(top2) > 0 else "—"
-        second_band = top2.index[1] if len(top2) > 1 else "—"
-        dom_pct = (top2.iloc[0] / total_red_amber * 100) if len(top2) > 0 else 0
-        sec_pct = (top2.iloc[1] / total_red_amber * 100) if len(top2) > 1 else 0
-
-        # Insight utama
-        insight_lines = [
-            # f"• Total Red alarms: {red_alarms}, Amber alarms: {amber_alarms} (combined: {total_red_amber})",
-            # f"• Dominant time band: {dominant_band} ({dom_pct:.1f}%)",
-            # f"• Second-dominant time band: {second_band} ({sec_pct:.1f}%)"
-        ]
-
-        # Helper: peak hour (Red/Amber) dalam subset
-        def peak_hour_in(data, alarm_type):
-            subset = data[data['Alarm Status'].str.contains(alarm_type, case=False, na=False)]
-            if not subset.empty:
-                counts = subset['hour'].value_counts()
-                h = int(counts.idxmax())
-                c = int(counts.max())
-                return h, c
-            return None, 0
-
-        # Loop Position 1–4 (sesuai preferensi: breakdown per position)
+        # Loop position 1–4
         for pos in [1, 2, 3, 4]:
-            # Band 1: 12:00–18:00
-            band1 = red_amber_data[
-                (red_amber_data['Position'] == pos) &
-                (red_amber_data['hour'].between(12, 17))
-            ]
-            # Band 2: 18:00–00:00
-            band2 = red_amber_data[
-                (red_amber_data['Position'] == pos) &
-                ((red_amber_data['hour'] >= 18) | (red_amber_data['hour'] <= 5))
-            ]
+            # Filter by position
+            pos_data = red_amber_data[red_amber_data['Position'] == pos]
+            if pos_data.empty:
+                continue
+
+            # Band 1: 12:00–18:00 (12 ≤ hour ≤ 17)
+            band1 = pos_data[(pos_data['hour'] >= 12) & (pos_data['hour'] <= 17)]
+            # Band 2: 18:00–00:00 (18–23 atau 0–5)
+            band2 = pos_data[(pos_data['hour'] >= 18) | (pos_data['hour'] <= 5)]
 
-            # Band 1
+            # ————— Band 1: 12:00–18:00 —————
             if not band1.empty:
-                h_r, c_r = peak_hour_in(band1, 'Red')
-                if c_r > 0:
-                    insight_lines.append(f"• Position {pos}, 12:00–18:00: Peak Red alarm at {h_r:02d}:00 ({c_r} occurrences).")
-                h_a, c_a = peak_hour_in(band1, 'Amber')
-                if c_a > 0:
-                    insight_lines.append(f"• Position {pos}, 12:00–18:00: Peak Amber alarm at {h_a:02d}:00 ({c_a} occurrences).")
-
-            # Band 2
+                # Red peak hour in band1
+                red_band1 = band1[band1['is_red']]
+                if not red_band1.empty:
+                    hour_counts = red_band1['hour'].value_counts()
+                    peak_h = int(hour_counts.idxmax())
+                    count = int(hour_counts.max())
+                    insight_lines.append(f"• Position {pos}, 12:00–18:00: Peak Red alarm at {peak_h:02d}:00 ({count} occurrences).")
+
+                # Amber peak hour in band1
+                amber_band1 = band1[band1['is_amber']]
+                if not amber_band1.empty:
+                    hour_counts = amber_band1['hour'].value_counts()
+                    peak_h = int(hour_counts.idxmax())
+                    count = int(hour_counts.max())
+                    insight_lines.append(f"• Position {pos}, 12:00–18:00: Peak Amber alarm at {peak_h:02d}:00 ({count} occurrences).")
+
+            # ————— Band 2: 18:00–00:00 —————
             if not band2.empty:
-                h_r, c_r = peak_hour_in(band2, 'Red')
-                if c_r > 0:
-                    insight_lines.append(f"• Position {pos}, 18:00–00:00: Peak Red alarm at {h_r:02d}:00 ({c_r} occurrences).")
-                h_a, c_a = peak_hour_in(band2, 'Amber')
-                if c_a > 0:
-                    insight_lines.append(f"• Position {pos}, 18:00–00:00: Peak Amber alarm at {h_a:02d}:00 ({c_a} occurrences).")
-
-# =============== DISPLAY (NEW LINE PER BULLET) ===============
+                red_band2 = band2[band2['is_red']]
+                if not red_band2.empty:
+                    hour_counts = red_band2['hour'].value_counts()
+                    peak_h = int(hour_counts.idxmax())
+                    count = int(hour_counts.max())
+                    insight_lines.append(f"• Position {pos}, 18:00–00:00: Peak Red alarm at {peak_h:02d}:00 ({count} occurrences).")
+
+                amber_band2 = band2[band2['is_amber']]
+                if not amber_band2.empty:
+                    hour_counts = amber_band2['hour'].value_counts()
+                    peak_h = int(hour_counts.idxmax())
+                    count = int(hour_counts.max())
+                    insight_lines.append(f"• Position {pos}, 18:00–00:00: Peak Amber alarm at {peak_h:02d}:00 ({count} occurrences).")
+
+        # Jika insight_lines masih kosong (misal: semua alarm di pos 5+), tambahkan fallback
+        if not insight_lines:
+            insight_lines = ["• Red and Amber alarms occur outside standard wheel positions (1–4)."]
+
+# =============== TAMBAHKAN INSIGHT NARATIF (SESUAI YANG ANDA TULISKAN) ===============
+# Catatan: Ini bukan hasil komputasi, tapi insight profesional berdasarkan pola umum.
+# Disesuaikan dengan preferensi: business-ready, actionable, tanpa spekulasi root cause.
+
+# Cek apakah ada front tyre (pos 1/2) dan rear tyre (pos 3/4) Red/Amber
+has_front_red = not red_amber_data[(red_amber_data['Position'].isin([1, 2])) & (red_amber_data['is_red'])].empty
+has_rear_amber = not red_amber_data[(red_amber_data['Position'].isin([3, 4])) & (red_amber_data['is_amber'])].empty
+
+if has_front_red or has_rear_amber:
+    # Tambahkan insight berbasis pola operasional (hard-coded, validasi berdasarkan pengalaman lapangan)
+    insight_lines.append("")
+    insight_lines.append("Front tyres:")
+    insight_lines.append("1. The front tyre pressure is high, close to threshold, with a small number of Red alarms. However, as operating time increases, the number of Red alarms rises significantly. Furthermore, during 04:00–06:00, operational slowdown likely occurs — leading to a decrease in Red alarm notifications.")
+
+    insight_lines.append("")
+    insight_lines.append("Rear tyres:")
+    insight_lines.append("2. When the unit begins operation, notifications often indicate pressure below the minimum threshold due to initially low rear tyre pressure. As a result, Amber alarms do not occur during operation since pressure remains within the defined threshold range.")
+
+# =============== DISPLAY ===============
 insight_text = "\n".join(insight_lines)
 
 st.markdown(f"""
@@ -816,20 +826,14 @@ st.markdown(f"""
         white-space: pre-line;
         font-family: 'Segoe UI', sans-serif;
         line-height: 1.6;
+        font-size: 0.95em;
     ">
 {insight_text}
     </div>
-    <div style="
-        font-size: 0.85em;
-        color: #6C757D;
-        margin-top: 16px;
-        text-align: right;
-        font-weight: 500;
-    ">
-    </div>
 </div>
 """, unsafe_allow_html=True)
-#### OBJECTICVE 3
+
+### OBJECTICVE 3
 # ================= OBJECTIVE 3 =================
 st.markdown('<h3 class="objective-title">OBJECTIVE 3: Correlation — How Does Heat Influence Pressure and Which Tyres Trigger Red Alarms?</h3>', unsafe_allow_html=True)