Research On Stock Index Prediction Model Based On PCA-IFOA-LSTM

Along with the rapid development of our economy and more mature development of the stock market,the status of the stock market in our economic system is constantly improving,which promotes people’s desire to invest in stock.But the rise and fall of the stock directly affect the interests of the majority of investors,but also to the social economy has a great fluctuation.Therefore,the research on the future trend of stock is particularly important.The trend of stock price is affected by political,macroeconomic and other factors,and the stock price series is nonlinear,noisy and non-stable,which makes it very difficult to predict the future trend of stock.The bionic algorithm’s optimization ability of weight threshold,the rapid development of deep learning in recent years and the excellent ability of nonlinear data processing have been applied by many scholars to the research of stock index prediction.Based on bionic algorithm and deep learning,this thesis explores the combination of algorithm and deep learning in stock index prediction.The main research is as follows:(1)Principal component analysis(PCA)was used to preprocess the acquired stock index data.Principal components were extracted from relevant technical indicators and basic data to reduce data dimensions,remove redundant data,and improve the learning rate of the neural network model.The main characteristic index is used to improve the overall prediction ability of the model.(2)The improved Drosophila Bionic optimization algorithm(IFOA)was introduced to obtain the optimal weights and thresholds of the LSTM neural network model.In order to improve the shortcomings of the Drosophila algorithm,the escape parameter is added to make the algorithm get the negative number,and the bacteria migration operation is added to make the algorithm quickly find the global optimal value,which improves the global search ability of the algorithm.Updating the weights and thresholds of the LSTM model can improve the training speed of the model,reduce the errors of network prediction and improve the accuracy.(3)The concept of PCA-IFOA-LSTM combined model was obtained by using long and short term neural network(LSTM)and related technologies.Not only the advantages of LSTM in time series data processing,but also the shortcomings of principal component analysis and bionic optimization algorithm are improved.By reducing the dimension of data features,the optimal weight threshold is obtained,and the long-term memory ability of LSTM neural network is further improved.This thesis compares three groups of models,including LSTM model,PCA-IFOA-RNN model and PCA-LSTM model.Through the prediction and comparison analysis with the three groups of comparison models,the combined model designed in this thesis is superior to the comparison model in every aspect.Both in the prediction accuracy,training speed and other aspects of the best performance.Through the calculation and comparison of MAE,RMSE and MAPE,the excellent performance of the combined model in this thesis is fully demonstrated,and the correctness and feasibility of the thought of predicting the future trend of the stock index in this thesis are proved.