Short-Term Predictions and LIME-Based Rule Extraction for Standard and Poor’s Index

• 作者: Chunqi Qi Wenjun Wu Xiuli Wang Yue Wang
• 来源期刊: 国际计算机前沿大会会议论文集
• 学科: 工学
• 关键词: Neural network Stock prediction LIME INTERPRETABILITY
• 年，卷（期）: 2020,（2）
• 页码范围: 329-343
• 页数: 15页
• 分类号: TP1

摘要

In this paper,neural networks is proposed to predict the trend of the S.P.500 index,comparing the effects of different data inputs and model types on the prediction results.Then the model was selected with good interpretation results to use the LIME interpretability algorithm to interpret the prediction results,and extract the prediction rules of the neural network model.It firstly compared three neural networks prediction models,including multi-layer perceptron(MLP),one-dimensional convolutional neural network(1D CNN)and long short-term memory(LSTM),with the price data,continuous technical indicators and discrete technical indicators of different time durations as inputs,and found the model and parameter configuration with good predicting effect for each type of inputs.In another aspect,the data of the first experiment was used,the rules were extracted in the model that predict the rise of the index by the LIME interpretability algorithm,and the rules with high investment utility were finally selected.The experimental result show high precision in predicting the trend of the stock index and high frequency of occurrence,with certain reference value for predicting the short-term index trend.