LR、LSTM、ARIMA、KNN等机器学习方法预测股价项目源码+文档说明

# 机器学习股价预测（使用多种模型实现） ## 文件说明 - forecast.py 数据处理、可视化、模型调用 - myutils.py 数据处理工具 - models.py 模型定义 - AMZN.csv 时间序列预测使用的数据集，为2010.01-2022.06的Amazon 股价变化情况，来自[Kaggle](https://www.kaggle.com/datasets/andrewmvd/sp-500-stocks?select=sp500_stocks.csv) - backtest.py 回测 - draw_line.py 回测结果展示，backtest.py中自动调用 - 各月股票队列.xlsx 回测时使用的股票队列，日期范围为2020.04-2020.12，来自参考项目2 - 个股持有情况分析.xlsx 回测结果文件 - 持有期收益率情况图.html 回测结果图，需要pyecharts打开 - 每日资金情况图.html 回测结果图，需要pyecharts打开 ## 环境配置 需要python3.8，其他依赖使用`pip install -r requirements.txt`命令安装 ## 运行说明 - 时间序列预测 运行命令```python forecast.py --model $modelname``` 运行时将`$modelname`替换为下面9种模型名称之一： 1. `linearRegression` 2. `DeterministProcess` 3. `RelativeStrengthIndex` 4. `ARIMA` 5. `DecisionTree` 6. `KNN` 7. `LSTM` 8. `Prophet` 9. `SVM` - 回测 运行命令```python backtest.py``` ## 实验结果 ``` linearRegression: Model train accuracy: 99.923% Model test accuracy: 98.442% Model train MAE: 0.424 Model train RMSE: 0.800 Model test MAE: 2.338 Model test RMSE: 3.252 DeterministProcess Model train accuracy: 99.923% Model test accuracy: 98.443% Model train MAE: 0.425 Model train RMSE: 0.799 Model test MAE: 2.339 Model test RMSE: 3.251 ARIMA Best model: ARIMA(0,1,0)(2,1,0)[12] Total fit time: 47.517 seconds Model test MAE: 73.689 Model test RMSE: 80.137 KNN Model test MAE: 47.711 Model test RMSE: 53.677 LSTM Model test MAE: 4.349 Model test RMSE: 5.701 Prophet Model test MAE: 39.901 Model test RMSE: 44.794 SVM accuracy=52.31% DecisionTree train accuracy: 0.550400 test accuracy: 0.476266 roc: 0.500000 RelativeStrengthIndex {'超买市场（RSI>=80）且实际下跌': 131, '超买市场（RSI>=80）但实际上涨': 127, '强势市场（50<=RSI<80）且实际下跌': 760, '强势市场（50<=RSI<80）但实际上涨': 858, '弱式市场（50>RSI>=20）且实际上涨': 640, '弱式市场（50>RSI>=20）但实际下跌': 547, '超卖市场（RSI<20）且实际上涨': 32, '超卖市场（RSI<20）但实际下跌': 22} ``` | Model | MAE | RMSE | |:------------------:|:------:|:------:| | linearRegression | 2.338 | 3.252 | | DeterministProcess | 2.339 | 3.251 | | ARIMA | 73.689 | 80.137 | | KNN | 47.711 | 53.677 | | LSTM | 4.349 | 5.701 | | Prophet | 39.901 | 44.794 | ## 参考项目 ### 纯机器学习预测（不涉及因子建模） 1. https://www.kaggle.com/code/nedahs/apple-stock-time-series-ml-models/notebook 2. https://github.com/moyuweiqing/A-stock-prediction-algorithm-based-on-machine-learning 3. https://github.com/LightingFx/hs300_stock_predict ### 多因子分析及预测（已放弃） - https://github.com/phonegapX/alphasickle - https://github.com/JoshuaWu1997/EMD-ALSTM-Multi-Factor-Stock-Profit-Prediction