liuian / independent_study_predict_stock_price_ncku Goto Github PK

原始資料：從investing.com抓csv檔；元大0050；Period: January 1, 2014, to November 4, 2023。

1. Goldencross:

   • Moving Average: Short = 5, Long = 20
   • Total Return Rate of Backtesting Analysis: 62.48%

2. Machine Learning-random forest:

   • 80% of the data is the training set, and 20% is the test set
   • R2 Score: 0.9997980640298204
   • Issue: No concept of time series

3. LSTM:

   • Using data from the past ten days as the training set for each training.
   • The first 80% of the data is the training set.
   • R2 Score: 0.9535430373937358

4. 綜合比較

   • 方法：Golden cross 以實際股市走勢回測、Random forest及LSTM以預測後的股價進行回測
   • Dataset日期區間：2014.01.02~2023.11.03
   • 預測區間：2021.11.10~2023-11-03
   • Golden Cress 長均線、短均線：5, 20

5. 結論：以下幾種情況都會影響交易的回報率

   • 不同時間區間
   • 不可以看到未來的資料random forest
   • 嘗試優化參數
   • Golden cross > random forest > random forest

標準化資料 scaler = MinMaxScaler(feature_range=(0, 1)) scaled_Data = scaler.fit_transform(Data) ... 反標準化預測結果 trainPredict = scaler.inverse_transform(trainPredict.reshape(-1, 1)) testPredict = scaler.inverse_transform(testPredict.reshape(-1, 1))

ValueError: non-broadcastable output operand with shape (1913,1) doesn't match the broadcast shape(1913,6) 因為scaler 同時紀錄了6個feature的轉換

改成 from sklearn.preprocessing import MinMaxScaler 假設Data是包含六個特徵的資料框 features = ['Price', 'Open', 'High', 'Low', 'Vol.', 'Change %'] X = Data[features] 初始化MinMaxScaler scalers = {} 對每個特徵獨立進行縮放 for feature in features: scaler = MinMaxScaler(feature_range=(0, 1)) X[feature] = scaler.fit_transform(X[[feature]]) scalers[feature] = scaler X現在包含獨立縮放後的每個特徵的值 X_array = X.values