基于改进的粒子群算法优化LSSVM股价预测研究

Study on the Prediction for Stock Price Based on the Optimized LSSVM of the Improved Particle Swarm Algorithm

为了提高股票价格的预测精度,针对股票价格数据的非平稳非线性的特性,本文运用改进的PSO实现LSSVM的核参数和惩罚系数自适应选择,提出一种SAPSO优化LSSVM股价预测模型,并以此进行实证分析。通过基于SAPSO-LSSVM算法的1步、3步、5步和7步预测结果和不同模型的预测时间和预测均方误差的对比结果可知,SAPSO-LSSVM股价预测模型具有预测精度高,预测时间短的优点,同时能够实现预测参数的自适应选择。

In order to improve the prediction accuracy of the stock price, stock price data for the nonlinear and non-stationary characteristics, this paper used the improved PSO to implement the self-adaptive selection of the LSSVM kernel parameter and penalty coefficient, and proposed a prediction model for stock price on SAPSO optimized LSSVM to analyze a case. The results showed that it had the high prediction accuracy, the advantages of short time and also could realize the self-adaptive selection for forecasting parameters based on prediction results in 1 step, 3 step, 5 step and the 7 on the SAPSO-LSSVM algorithm and the comparison between prediction time and the mean square error of different models.