Research On ICA Methods Of Stock Return Forecasting

Stock return modeling is an important research field in econometrics.Multivariate volatility model provides a powerful tool for maximizing asset benefits and minimizing risk through portfolio and asset allocation.In multivariate volatility model,ICA algorithm can achieve effective dimensionality reduction,but it also has the disadvantages of low convergence accuracy and the problem that it is easy to fall into local optimum.In this paper,PSO algorithm and HHO algorithm are used to improve ICA algorithm,and two types of new multivariate volatility models are proposed.One is PSO-ICA-GARCH model,the other is HHO-ICA-EGARCH and HHO-ICA-TGARCH models,which are respectively used for the empirical research on the prediction of the stock return volatility with different leverage effects.The results show that these three models have better model fitting results and more accurate model prediction results.The main work of this paper is as follows.(1)Research on the PSO-ICA-GARCH model of stock income fluctuation rate.The PSO-ICA algorithm is an optimization algorithm based on negative entropy maximization.The global optimization ability of PSO algorithm makes the algorithm not easily fall into the local optimal value.Compared with the traditional ICA,PSO-ICA algorithm has higher separation accuracy.Based on the PSO-ICA algorithm,the PSO-ICA-GARCH model is established,which effectively solves the problem that the prediction effect is not consistent with the reality in O-GARCH model when the principal components are weakly correlated.Since GARCH model affects future fluctuation rates through historical volatility,the PSO-ICA-GARCH model is suitable for modeling stock return series with weak leverage effect.In the empirical analysis,part of Alibaba concept stocks are tested,and the autocorrelation of the selected stock yield data is analyzed.Then the ARCH effect test is carried out on the return sequence of each stock,and the sequence without ARCH effect is eliminated.Finally,the PSO-ICA-GARCH model is established to predict the return volatility.The results show that the PSO-ICA-GARCH model has better model fitting effect and more accurate model prediction effect than the O-GARCH and ICA-GARCH models.(2)Research on HHO-ICA-EGARCH and HHO-ICA-TGARCH Models Prediction of stock income fluctuation rate.In this paper,HHO algorithm based on population,natural heuristic and gradient free optimization is used to improve ICA algorithm,and a new HHO-ICA algorithm based on negative entropy maximization is presented.This algorithm can also solve the problem that it is easy to fall into local optimal value in the traditional ICA,and has higher separation accuracy compared with PSO-ICA algorithm.Both EGARCH models and TGARCH models can better explain the strong leverage effects of securities market asset yield sequence.In this paper,HHO-ICA is introduced into these two models,and then two new multivariate volatility models,HHO-ICA-EGARCH and HHO-ICA-TGARCH are proposed,and applied to the modeling process of stock return volatility with strong stick effect.Some stock data of science and technology innovation board are used for empirical analysis.The results show that,compared with ICA-EGARCH and PSO-ICA-EGARCH model,the HHO-ICA-EGARCH model has better model fitting effect and more accurate 5-step backward prediction effect,which also indicates that HHO-ICA-EGARCH model is more suitable for modeling return series with strong leverage effect than PSO-ICA-EGARCH model.Compared with ICA-TGARCH model,PSO-ICA-TGARCH and HHO-ICA-TGARCH models have better model fitting effect.In terms of the prediction effect,the HHO-ICA-TGARCH model has a better 5-step backward prediction effect than PSO-ICA-TGARCH and ICA-TGARCH models,which indicates that the HHO-ICA-TGARCH model is also suitable for the return series with strong leverage effect.Both HHO-ICA-EGARCH and HHO-ICA-TGARCH models provide useful tools for accurate prediction of stock return volatility.