Upload model_predictor.py

model_predictor.py

@@ -0,0 +1,152 @@
+# model_predictor.py
+
+import numpy as np
+import pandas as pd
+from tensorflow.keras.models import load_model
+import joblib
+import yfinance as yf
+
+# --- 模型與設定檔 (未來訓練好後，請將檔案放在同目錄下) ---
+MODEL_FILE = 'stock_predictor_model.h5'
+SCALER_X_FILE = 'scaler_X.pkl'
+SCALER_Y_FILE = 'scaler_y.pkl'
+LOOKBACK_DAYS = 30 # 必須與訓練時的 LOOKBACK_DAYS 相同
+
+# --- 啟動時載入模型與縮放器 (只會載入一次) ---
+try:
+    model = load_model(MODEL_FILE)
+    scaler_X = joblib.load(SCALER_X_FILE)
+    scaler_y = joblib.load(SCALER_Y_FILE)
+    print("進階 LSTM 模型與縮放器載入成功。")
+except Exception as e:
+    print(f"提示：未找到或無法載入進階模型檔案 ({e})。應用將使用簡易統計模型。")
+    model = None
+
+# --- 從 app.py 複製過來的技術指標計算函式 ---
+# 確保資料準備的邏輯一致
+def calculate_technical_indicators(df):
+    """計算技術指標"""
+    if df.empty: return df
+    df['MA5'] = df['Close'].rolling(window=5).mean()
+    df['MA20'] = df['Close'].rolling(window=20).mean()
+    delta = df['Close'].diff()
+    gain = (delta.where(delta > 0, 0)).rolling(window=14).mean()
+    loss = (-delta.where(delta < 0, 0)).rolling(window=14).mean()
+    rs = gain / loss
+    df['RSI'] = 100 - (100 / (1 + rs))
+    exp1 = df['Close'].ewm(span=12).mean()
+    exp2 = df['Close'].ewm(span=26).mean()
+    df['MACD'] = exp1 - exp2
+    df['MACD_Signal'] = df['MACD'].ewm(span=9).mean()
+    df['MACD_Histogram'] = df['MACD'] - df['MACD_Signal']
+    low_min = df['Low'].rolling(window=9).min()
+    high_max = df['High'].rolling(window=9).max()
+    rsv = (df['Close'] - low_min) / (high_max - low_min) * 100
+    df['K'] = rsv.ewm(com=2).mean()
+    df['D'] = df['K'].ewm(com=2).mean()
+    df['up_move'] = df['High'] - df['High'].shift(1)
+    df['down_move'] = df['Low'].shift(1) - df['Low']
+    df['+DM'] = np.where((df['up_move'] > df['down_move']) & (df['up_move'] > 0), df['up_move'], 0)
+    df['-DM'] = np.where((df['down_move'] > df['up_move']) & (df['down_move'] > 0), df['down_move'], 0)
+    df['TR'] = np.max([df['High'] - df['Low'], abs(df['High'] - df['Close'].shift(1)), abs(df['Low'] - df['Close'].shift(1))], axis=0)
+    df['+DI'] = (df['+DM'].ewm(com=13, adjust=False).mean() / df['TR'].ewm(com=13, adjust=False).mean()) * 100
+    df['-DI'] = (df['-DM'].ewm(com=13, adjust=False).mean() / df['TR'].ewm(com=13, adjust=False).mean()) * 100
+    df['DX'] = abs(df['+DI'] - df['-DI']) / (df['+DI'] + df['-DI']) * 100
+    df['ADX'] = df['DX'].ewm(com=13, adjust=False).mean()
+    return df
+
+
+def get_all_features_for_model(period="3y"):
+    """獲取並整合模型需要的所有15個特徵。"""
+    print("正在下載市場數據以準備進階模型輸入...")
+    tickers = {'^TWII': 'TWII', '^SOX': 'SOX', 'TSM': 'TSM_ADR'}
+    data_yf = yf.download(list(tickers.keys()), period=period, auto_adjust=True)
+    
+    twii_data = data_yf.loc[:, ('Open', 'High', 'Low', 'Close', 'Volume')]['TWII'].copy()
+    
+    print("正在計算技術指標...")
+    df_main = calculate_technical_indicators(twii_data)
+    
+    print("正在合併外部市場與檔案數據...")
+    df_main['費城 半導體'] = data_yf['Close']['SOX']
+    df_main['台積電 ADR'] = data_yf['Close']['TSM_ADR']
+    
+    try:
+        df_climate = pd.read_csv('business_climate.csv')
+        df_climate['Date'] = pd.to_datetime(df_climate['Date'].astype(str) + '-01')
+        df_climate = df_climate.set_index('Date').rename(columns={'Index': '景氣燈號'})
+        
+        df_pmi = pd.read_csv('taiwan_pmi.csv')
+        df_pmi['Date'] = pd.to_datetime(df_pmi['DATE'].astype(str) + '-01')
+        df_pmi = df_pmi.set_index('Date').rename(columns={'INDEX': 'PMI'})
+        
+        df_main = pd.merge(df_main, df_climate, left_index=True, right_index=True, how='left')
+        df_main = pd.merge(df_main, df_pmi, left_index=True, right_index=True, how='left')
+    except FileNotFoundError as e:
+        print(f"警告: 找不到檔案 {e.filename}，相關欄位將為空。")
+        df_main['景氣燈號'] = np.nan
+        df_main['PMI'] = np.nan
+
+    df_main.fillna(method='ffill', inplace=True)
+    df_main.dropna(inplace=True)
+
+    df_final = df_main.rename(columns={
+        'Close': '加權指數', 'Volume': '成交量', 'K': 'K線', 'D': 'D線',
+        'MACD_Signal': 'MACD信號線', 'MACD_Histogram': 'MACD柱狀圖'
+    })
+    
+    print("所有特徵整合完畢！")
+    return df_final
+
+
+def advanced_lstm_predict(predict_days: int = 5):
+    """使用訓練好的 LSTM 模型進行預測。"""
+    if model is None:
+        print("進階模型未載入，無法進行預測。")
+        return None
+
+    try:
+        # 1. 獲取並整合所有最新資料
+        all_features_df = get_all_features_for_model()
+        if len(all_features_df) < LOOKBACK_DAYS:
+            print("��料長度不足，無法進行進階預測。")
+            return None
+            
+        # 2. 準備輸入資料
+        FEATURES = [
+            '加權指數', '成交量', '費城 半導體', '台積電 ADR', 'RSI', 'MACD', 
+            'MACD信號線', 'MACD柱狀圖', 'K線', 'D線', '+DI', '-DI', 'ADX', 
+            '景氣燈號', 'PMI'
+        ]
+        
+        last_sequence_df = all_features_df[FEATURES].tail(LOOKBACK_DAYS)
+        
+        if last_sequence_df.isnull().values.any():
+            print("警告：輸入的序列資料中存在缺失值，無法預測。")
+            return None
+            
+        input_scaled = scaler_X.transform(last_sequence_df)
+        input_reshaped = np.reshape(input_scaled, (1, LOOKBACK_DAYS, len(FEATURES)))
+        
+        # 3. 執行預測
+        prediction_scaled = model.predict(input_reshaped)
+        prediction_unscaled = scaler_y.inverse_transform(prediction_scaled)
+        
+        # 4. 處理預測結果
+        target_map = {1: 0, 5: 1, 10: 2}
+        if predict_days not in target_map:
+            predict_days = 5 # 預設值
+            
+        predicted_price = prediction_unscaled[0][target_map[predict_days]]
+        
+        last_price = all_features_df['加權指數'].iloc[-1]
+        change_pct = ((predicted_price - last_price) / last_price) * 100
+        
+        return {
+            'predicted_price': predicted_price,
+            'change_pct': change_pct,
+            'confidence': 0.85 # 可設為固定值
+        }
+    except Exception as e:
+        print(f"執行進階預測時發生錯誤: {e}")
+        return None