Research On The Combination Model Of Stock Price Index Prediction Based On Multi-objective Antlion Optimization Algorithm And Neural Network

Mastering the development laws of the stock market and providing a good financial investment environment are necessary prerequisites for the healthy development of the stock market.Therefore,accurate and effective forecasting of stock price indices has gradually become a research focus in the financial field.However,the stock system is vulnerable to many factors owing to its complexity and nonlinearity,that makes the forecasting work especially intricate and difficult.Based on the daily K-line data of stock indexes,combined with neural network methods and optimization algorithms,this paper develops a stock index price trend prediction model,comprehensively analyzes and mines stock information,and predicts and analyzes future stock index prices.The following aspects are the main summary of this paper:First,this article applies a new signal processing technology-Complementary Ensemble Mode Noise Decomposition(CEEMD)to preprocess the stock index time series,decompose and reorganize the original data,remove the noise part of historical stock price data to retain its main information;then utilize autoregressive moving average(ARIMA)model of statistical model and three classic neural network models,namely,error back propagation neural network(BPNN),Elman neural network(ENN),wavelet neural network(WNN),to respectively predict the closing price of future stock indexes;next,according to the idea of the combined model,the Multi-Objective Antlion Optimization algorithm(MOALO)in the optimization algorithm is selected to combine the prediction results of each single model.Finally,the predicted values of the combined model are output.This method can effectively integrate the advantages of each model and make up for the defects of each model;then,the accuracy and stability of the model are verified by mean absolute percentage error(MAPE),mean absolute error(MAE),root mean error(RMSE)and sum of squares(SSE);finally,the effectiveness of the model is measured and the comprehensive analysis is made based on the K-line theory.This paper selects the model's effective measurement to achieve a comparative analysis of model effectiveness,and compares the combination strategy of multi-objective ant lion optimization algorithm and the simple combination average combination strategy in each error.Based on the above analysis,this article draws the following conclusions:first of all,the CEEMD-MOALO combination model proposed in this paper has the highest accuracy and the best effect in each stock index prediction,and its predicted value and the true closing of the stock index The degree of fit between prices is higher.According to the drawn scatter plot between the predicted value and the true value,it can be observed that the graph is closer to a straight line passing through the origin.The probability distribution shows that the prediction error obeys the normal distribution.The results of several experiments show that the CEEMD-MOALO combination model with highest stability can maintain the optimal prediction accuracy.In the aspect of combination strategy,this paper compares the combination model with the simple average model in MAPE,MAE,RMSE and other error indicators.It can be found that the CEEMD-MOALO combination model has a total of 5 stock price indexes and 4 error indicators.In the 20 verifications,good prediction performance was shown,which indicates that the combination strategy of the CEEMD-MOALO combination model is more superior than the general combination average;combined with the trend graph and K-line graph analysis of the prediction results,through comparative analysis,the combination model shows high sensitivity at the turning point of the trend,and the trend reversal analysis can remind investors to adjust their investment strategy,which is the practical application value of this portfolio model.To sum up,the advantage of the CEEMD-MOALO combination model in this paper is not only to overcome the limitations of traditional technical indicators,but also makes up for the inherent shortcomings of a single model,as well as the lack of model stability and effectiveness.The comprehensive analysis performance of stock forecasting system is optimized.Finally,the proposed model has strong practicability.In order to avoid the influence of time period on the forecasting effect,this paper verifies that the proposed model is more practical than traditional models for forecasting the market index in different periods.