alphanet 0.0.4

A recurrent neural network for predicting stock market performance

AlphaNetV3

A Recurrent Neural Network For Predicting Stock Prices

Requirement: tensorflow 2.

1. The model and data utility are in ./alphanet/ folder.
2. CSI500 and CSI800 stock market data are included in ./data/ folder.
3. Model Configuration is in config.py.
4. To run the model, execute python main.py.
5. The loss plot pictures and models will be stored by each epoch and training period in ./models/ folder.
6. Tests of layers and time series data tools are in ./tests/ folder. To run the test, execute python -m unittest tests.tests.

API Overview

Modules

• alphanet: 复现华泰金工 alpha net V3 版本.
• alphanet.data: 多维多时间序列神经网络滚动训练数据工具箱.
• alphanet.metrics: 训练的计量信息.

Classes

• alphanet.AlphaNetV3: alpha net v3版本模型.
• alphanet.Correlation: 每个stride各个时间序列的相关系数.
• alphanet.Covariance: 每个stride各个时间序列的covariance.
• alphanet.FeatureExpansion: 时间序列特征扩张层.
• alphanet.LinearDecay: 每个序列各个stride的线性衰减加权平均.
• alphanet.Return: 每个序列各个stride的回报率.
• alphanet.Std: 每个序列各个stride的标准差.
• alphanet.ZScore: 每个序列各个stride的均值除以其标准差.
• data.TimeSeriesData: 单个时间序列信息.
• data.TrainValData: 该类用于生成不同训练阶段的tensorflow dataset.
• metrics.UpDownAccuracy: 通过对return的预测来计算涨跌准确率.

Functions

• alphanet.load_model: 包装tf.kreas的load_model，添加UpDownAccuracy.

---

This file was automatically generated via lazydocs.