Update app.py

app.py

@@ -1031,6 +1031,7 @@ try:
     from wordcloud import WordCloud
     import matplotlib.pyplot as plt
     import plotly.express as px
+    from collections import Counter
     WORDCLOUD_AVAILABLE = True
 except ImportError:
     WORDCLOUD_AVAILABLE = False
@@ -1082,36 +1083,33 @@ if WORDCLOUD_AVAILABLE:
                 pull_values = []
                 for cat in category_counts.index:
                     if cat in ['Unsafe Action', 'Near Miss']:
-                        pull_values.append(0.1)  # Tarik keluar sedikit
+                        pull_values.append(0.1)
                     else:
                         pull_values.append(0.0)
 
                 fig_pie.update_traces(
-                    textposition='outside',  # Label di luar slice
-                    textinfo='percent+label',  # Tampilkan persentase + nama
-                    pull=pull_values,          # Tarik slice keluar untuk Unsafe Action & Near Miss
+                    textposition='outside',
+                    textinfo='percent+label',
+                    pull=pull_values,
                     marker=dict(line=dict(color='#FFFFFF', width=1))
                 )
                 fig_pie.update_layout(
-                    showlegend=False,  # ❌ matikan legend default
+                    showlegend=False,
                     height=450,
                     margin=dict(t=20, b=20, l=40, r=40)
                 )
                 st.plotly_chart(fig_pie, use_container_width=True)
 
-                # ✅ CUSTOM LEGEND: Semua kategori dengan ikon GARIS HORIZONTAL kecil (mirip referensi)
+                # ✅ CUSTOM LEGEND
                 st.markdown("<div style='text-align: center; font-size: 14px; margin-top: -10px;'>", unsafe_allow_html=True)
                 legend_items = []
-
                 for cat in category_counts.index:
                     color = color_map.get(cat, '#9E9E9E')
                     item = (
                         f"<span style='display: inline-block; width: 16px; height: 4px; background-color: {color}; "
-                        f"margin: 0 8px; border-radius: 1px;'></span>"
-                        f"{cat}"
+                        f"margin: 0 8px; border-radius: 1px;'></span>{cat}"
                     )
                     legend_items.append(item)
-
                 legend_html = " | ".join(legend_items)
                 st.markdown(f"<div>{legend_html}</div>", unsafe_allow_html=True)
                 st.markdown("</div>", unsafe_allow_html=True)
@@ -1137,15 +1135,16 @@ if WORDCLOUD_AVAILABLE:
             st.warning("No data available after filtering out 'Positive' category.")
         else:
             if 'keyword_kategori' in df_filtered_kategori.columns:
-                # Gabungkan text
+                # Gabungkan & bersihkan teks
                 import re
-                text = ' '.join(df_filtered_kategori['keyword_kategori'].dropna().astype(str))
-                text = re.sub(r'[^a-zA-Z\s]', ' ', text)
+                text_series = df_filtered_kategori['keyword_kategori'].dropna().astype(str)
+                text = ' '.join(text_series)
+                text = re.sub(r'[^a-zA-Z\s]', ' ', text).strip()
 
-                if text.strip():
-                    # Buat WordCloud
+                if text:
+                    # Generate wordcloud
                     wordcloud = WordCloud(
-                        width=1600,          # resolusi besar agar HD
+                        width=1600,
                         height=800,
                         background_color='white',
                         colormap='viridis',
@@ -1153,13 +1152,30 @@ if WORDCLOUD_AVAILABLE:
                         random_state=42
                     ).generate(text)
 
-                    # Output kecil tapi tajam
+                    # Tampilkan
                     fig, ax = plt.subplots(figsize=(3, 2), dpi=200)
                     ax.imshow(wordcloud, interpolation='bilinear')
                     ax.axis('off')
                     plt.tight_layout()
-
                     st.pyplot(fig, use_container_width=True)
+
+                    # 🔍 🔹 Insight dinamis: top 3 keywords
+                    words_clean = [w.lower() for w in text.split() if len(w) > 2]
+                    top3 = [word.capitalize() for word, _ in Counter(words_clean).most_common(3)]
+                    if top3:
+                        insight = f"The most frequent unsafe issue themes are: <strong>{', '.join(top3)}</strong>."
+                    else:
+                        insight = "Recurring unsafe issue patterns are detectable in textual findings."
+
+                    st.markdown(
+                        f"""
+                        <div style='text-align: center; font-size: 14px; color: #2c3e50; margin-top: 12px; line-height: 1.5;'>
+                             <strong>Insight:</strong> {insight}
+                        </div>
+                        """,
+                        unsafe_allow_html=True
+                    )
+
                 else:
                     st.warning("No valid text remaining after cleaning.")
             else: