Corrigir api

predict.py

@@ -88,8 +88,14 @@ class DenguePredictor:
         municipio_name = self.municipios[self.municipios['codigo_ibge'] == ibge_code].iloc[0]['municipio']
 
         dynamic_features = list(self.feature_names_pt.keys())
-        last_real_sequence = df_mun[dynamic_features].iloc[-self.sequence_length:].copy()
-        dynamic_sequence_scaled = scaler_dyn.transform(last_real_sequence)
+        # Pegar últimas 'sequence_length' semanas para iniciar a previsão
+        last_sequence = df_mun[dynamic_features].iloc[-self.sequence_length:].copy()
+
+        # Substituir NaNs da sequência inicial por zeros apenas para o scaler (não para salvar)
+        # Isso é só para evitar erro do scaler; depois o modelo vai gerar previsão real
+        last_sequence_filled = last_sequence.fillna(0)
+        dynamic_sequence_scaled = scaler_dyn.transform(last_sequence_filled)
+
         static_data = df_mun[["latitude", "longitude"]].iloc[0].values.reshape(1, -1)
         static_input = scaler_static.transform(static_data)
 
@@ -115,7 +121,8 @@ class DenguePredictor:
             if (future_year, future_week) in climate_lookup:
                 future_climate = climate_lookup[(future_year, future_week)]
             else:
-                future_climate = df_mun[dynamic_features[1:]].tail(4).mean()  # só clima, sem numero_casos
+                future_climate = df_mun[dynamic_features[1:]].tail(4).mean(axis=0, skipna=True)
+
 
             new_row = np.zeros(len(dynamic_features), dtype=np.float32)
             new_row[0] = pred_scaled  # numero_casos previsto
@@ -128,19 +135,28 @@ class DenguePredictor:
 
         # Histórico das últimas 52 semanas
         historic_data = [
-            {"date": row['data_semana_iso'].strftime('%Y-%m-%d'), "cases": row["numero_casos"]}
+            {"date": row['data_semana_iso'].strftime('%Y-%m-%d'),
+            "cases": int(row["numero_casos"]) if not pd.isna(row["numero_casos"]) else 0}
             for _, row in df_mun.tail(52).iterrows()
         ]
 
+
         # Análise de lag
         df_analysis = df_mun[dynamic_features].rename(columns=self.feature_names_pt)
         max_lag = 12
         cases_col_name = 'Nº de Casos de Dengue'
         lag_features = ['Temperatura Média (°C)', 'Precipitação (mm)']
-        lag_correlations = {
-            col: [df_analysis[cases_col_name].corr(df_analysis[col].shift(lag)) for lag in range(1, max_lag + 1)]
-            for col in lag_features
-        }
+        lag_correlations = {}
+        for col in lag_features:
+            corrs = []
+            for lag in range(1, max_lag + 1):
+                x = df_analysis[col].shift(lag)
+                y = df_analysis[cases_col_name]
+                valid_idx = x.notna() & y.notna()
+                corr = y[valid_idx].corr(x[valid_idx]) if valid_idx.any() else 0
+                corrs.append(corr)
+            lag_correlations[col] = corrs
+
 
         plt.figure(figsize=(10, 6), facecolor='#18181b')
         ax = plt.gca()