Update app.py

app.py

@@ -58,38 +58,35 @@ def calcular_fator_avaliacao(titulo, EC, PU):
     fator_avaliacao = max((4 * ec_pontuacao + 6 * PVU - 3 * PUB) / 100, VR)
     return fator_avaliacao
 
-def select_nearest_items(df, query):
-    # Calculate title similarity for relevance
+def select_nearest_items_adjusted(df, query):
+    # Lower the title similarity threshold if necessary
     df['Title_Similarity'] = df['Title'].apply(lambda x: fuzz.WRatio(query, x))
-    # Filter by title similarity threshold
-    df_filtered_by_similarity = df[df['Title_Similarity'] > 70]  # Adjust similarity threshold as needed
-    
-    if df_filtered_by_similarity.empty:
-        # If no closely matching titles are found, return an empty DataFrame
-        return pd.DataFrame()
-    
-    # Filter by price, considering only those within a reasonable range
-    reasonable_price_df = df_filtered_by_similarity[df_filtered_by_similarity['Price'] <= df_filtered_by_similarity['Price'].quantile(0.75)]
-    # Calculate distance from target price for sorting
-    target_price = reasonable_price_df['Price'].mode().min() if not reasonable_price_df['Price'].mode().empty else reasonable_price_df['Price'].median()
-    reasonable_price_df['Distance'] = (reasonable_price_df['Price'] - target_price).abs()
+    df_filtered = df[df['Title_Similarity'] > 50]  # Adjusted threshold
+
+    # Calculate mode price in a more inclusive manner
+    mode_price = df_filtered['Price'].mode()
+    if mode_price.empty:
+        target_price = df_filtered['Price'].median()
+    else:
+        target_price = mode_price.min()
 
-    # Initialize container for selected items ensuring different marketplaces
-    selected_items = []
-    included_marketplaces = set()
+    df_filtered['Distance'] = (df_filtered['Price'] - target_price).abs()
+    df_sorted = df_filtered.sort_values(['Distance', 'Title_Similarity'], ascending=[True, False])
 
-    # Sort by distance to target price, then by title similarity
-    df_sorted = reasonable_price_df.sort_values(['Distance', 'Title_Similarity'], ascending=[True, False])
+    # Ensure diversity in marketplaces
+    marketplaces_selected = set()
+    results = []
 
     for _, row in df_sorted.iterrows():
-        marketplace = row['Marketplace']
-        if marketplace not in included_marketplaces:
-            selected_items.append(row)
-            included_marketplaces.add(marketplace)
-        if len(selected_items) >= 5:
+        if row['Marketplace'] not in marketplaces_selected and len(marketplaces_selected) < 3:
+            results.append(row)
+            marketplaces_selected.add(row['Marketplace'])
+
+        if len(results) >= 3:
             break
 
-    return pd.DataFrame(selected_items)
+    return pd.DataFrame(results)
+
 
 def search_with_fallback(query, df, limit=15):
     query_parts = query.split()