fix CustomNextPlaceModel.py

CustomNextPlaceModel.py

@@ -126,36 +126,50 @@ class CustomNextPlaceModel:
         combined_dataset, _ = dp.create_convolution_features(combined_dataset, combined_dataset.columns.to_list(), 3)
 
         # Predict B scores for different categories
-        score_B_1 = self.score_b_1.predict_proba_dataset(combined_dataset[combined_dataset['A']==1])
-        score_B_2 = self.score_b_2.predict_proba_dataset(combined_dataset[combined_dataset['A']==2])
-        score_B_3 = self.score_b_3.predict_proba_dataset(combined_dataset[combined_dataset['A']==3])
+        # score_B_1 = self.score_b_1.predict_proba_dataset(combined_dataset[combined_dataset['A']==1])
+        # score_B_2 = self.score_b_2.predict_proba_dataset(combined_dataset[combined_dataset['A']==2])
+        # score_B_3 = self.score_b_3.predict_proba_dataset(combined_dataset[combined_dataset['A']==3])
+        b_scores = {
+            '1': self.score_b_1.predict_proba_dataset(combined_dataset[combined_dataset['A'] == 1])
+            if not combined_dataset[combined_dataset['A'] == 1].empty else pd.DataFrame(
+                {'B_Probability_Class_0': [0], 'B_Probability_Class_1': [0], 'B_Probability_Class_2': [0]}),
+            '2': self.score_b_2.predict_proba_dataset(combined_dataset[combined_dataset['A'] == 2])
+            if not combined_dataset[combined_dataset['A'] == 2].empty else pd.DataFrame(
+                {'B_Probability_Class_0': [0], 'B_Probability_Class_1': [0], 'B_Probability_Class_2': [0]}),
+            '3': self.score_b_3.predict_proba_dataset(combined_dataset[combined_dataset['A'] == 3])
+            if not combined_dataset[combined_dataset['A'] == 3].empty else pd.DataFrame(
+                {'B_Probability_Class_0': [0], 'B_Probability_Class_1': [0], 'B_Probability_Class_2': [0]}),
+        }
 
         # Concatenate B scores
-        df_B = pd.concat([score_B_1, score_B_2, score_B_3], ignore_index=True)
+        df_B = pd.concat([b_scores['1'], b_scores['2'], b_scores['3']], ignore_index=True)
+
+        df_B_ = df_B.dropna()
 
         # Further combine and process dataset
-        combined_dataset = dp.combine_datasets(df_B, dp.X)
+        combined_dataset = dp.combine_datasets(df_B_, dp.X)
         combined_dataset = combined_dataset.drop(columns=['0'])
         combined_dataset, _ = dp.create_convolution_features(combined_dataset, combined_dataset.columns.to_list(), 3)
 
         # Predict C scores for different categories
         c_scores = {
             '1': self.score_c_models['1'].predict_dataset(combined_dataset[combined_dataset['B'].isin([1])])
-                  if not combined_dataset[combined_dataset['B'].isin([1])].empty else pd.DataFrame({'price': [0]}),
+            if not combined_dataset[combined_dataset['B'].isin([1])].empty else pd.DataFrame({'price': [0]}),
             '2': self.score_c_models['2'].predict_dataset(combined_dataset[combined_dataset['B'].isin([2])])
-                  if not combined_dataset[combined_dataset['B'].isin([2])].empty else pd.DataFrame({'price': [0]}),
+            if not combined_dataset[combined_dataset['B'].isin([2])].empty else pd.DataFrame({'price': [0]}),
             '3_4': self.score_c_models['3_4'].predict_dataset(combined_dataset[combined_dataset['B'].isin([3, 4])])
-                  if not combined_dataset[combined_dataset['B'].isin([3, 4])].empty else pd.DataFrame({'price': [0]}),
+            if not combined_dataset[combined_dataset['B'].isin([3, 4])].empty else pd.DataFrame({'price': [0]}),
             '5_6': self.score_c_models['5_6'].predict_dataset(combined_dataset[combined_dataset['B'].isin([5, 6])])
-                  if not combined_dataset[combined_dataset['B'].isin([5, 6])].empty else pd.DataFrame({'price': [0]}),
+            if not combined_dataset[combined_dataset['B'].isin([5, 6])].empty else pd.DataFrame({'price': [0]}),
             '7': self.score_c_models['7'].predict_dataset(combined_dataset[combined_dataset['B'].isin([7])])
-                  if not combined_dataset[combined_dataset['B'].isin([7])].empty else pd.DataFrame({'price': [0]}),
+            if not combined_dataset[combined_dataset['B'].isin([7])].empty else pd.DataFrame({'price': [0]}),
             '8_9': self.score_c_models['8_9'].predict_dataset(combined_dataset[combined_dataset['B'].isin([8, 9])])
-                  if not combined_dataset[combined_dataset['B'].isin([8, 9])].empty else pd.DataFrame({'price': [0]})
+            if not combined_dataset[combined_dataset['B'].isin([8, 9])].empty else pd.DataFrame({'price': [0]})
         }
         df_C = pd.concat(
             [c_scores[key][['price']] for key in c_scores
-            if isinstance(c_scores[key], pd.DataFrame) and 'price' in c_scores[key].columns and not c_scores[key].empty],
+             if
+             isinstance(c_scores[key], pd.DataFrame) and 'price' in c_scores[key].columns and not c_scores[key].empty],
             ignore_index=True
         )
 
@@ -178,12 +192,12 @@ class CustomNextPlaceModel:
         result = self.predict(input_data)
         predicted_sale_price, predicted_days = result['price'].iloc[0], result['days'].iloc[0] # кол-во дней нужно преобразовать в дату в виде строки
 
-        current_days_on_market = input_data.get('days_on_market', 0) or 0
+        current_days_on_market = input_data['days_on_market'].iloc[0] if 'days_on_market' in input_data else 0
 
         # Вычисление даты размещения на рынке
-        date_listed = datetime.now() - timedelta(days=current_days_on_market)
+        date_listed = datetime.now() - timedelta(days=int(current_days_on_market))
 
         # Вычисление предсказанной даты продажи
-        predicted_sale_date = (date_listed + timedelta(days=predicted_days)).strftime('%Y-%m-%d')
+        predicted_sale_date = (date_listed + timedelta(days=int(predicted_days))).strftime('%Y-%m-%d')
 
-        return predicted_sale_price, predicted_sale_date
+        return float(predicted_sale_price), predicted_sale_date