股票预测中AP-FIG和DKIPSO-SVR模型的建立与应用

Construction and Application of AP-FIG and DKIPSO-SVR Models for Stock Forecast

针对股市指标因子间存在高冗余、非线性等特点,同时开盘价还受政治、经济、投资者心理等因素影响,使得线性方法无法准确地预测开盘价的走势及波动范围.因此,为了提高开盘价预测的准确率,首次提出AP-FIG算法处理收盘价,同时首次在动态改进的粒子群算法(DIPSO)位置更新过程中引入缩减因子,并对其动态调整(DKIPSO),构建基于DKIPSO自动选取SVR参数的开盘价预测模型(DKIPSO-SVR).仿真结果表明,相比于DIPSO-SVR,DKIPSO-SVR不仅提高了泛化能力,还增强了稳健性;对比于GS-SVR,DKIPSO-SVR不仅误差低、拟合效果好,同时还将精度提高近7个百分点.最重要的是,利用AP-FIG算法处理收盘价的预测结果优于其他隶属度方程,使得AP-FIG和DKIPSO-SVR组合模型在股市开盘价预测方面具有更加广泛的应用价值.

The index factor of stock market has characters such as nonlinear and high redundan- cy,meanwhile,the opening price is also affected by policy, economy, psychology of investors and other factors. Therefore,it is difficult for linear methods to forecast the trend and fluctuation of the opening price accurately. In order to improve the prediction accuracy of the stock market opening price,for the first time, the AP-FIG algorithm is proposed to process the closing price samples, meanwhile,a DKIPSO model is proposed as the reduction factor and is introduced to the dynamic improved Particle Swarm Optimization （DIPSO） when updating the location to dynamic adjust; on the basis of previous model,DKIPSO-SVR model is established. Simulation results show that com- pared to the DIPSO-SVR model, the proposed DKIPSO-SVR model has higher precision and better robustness; and compared to the GS-SVR model, the DKIPSO-SVR model not only has low error and better fitting results, but also improves the evaluation accuracy by 7 %. It is most important that the DKIPSO-SVR model employs AP-FIG algorithm to forecast the effect of closing price and a- chieves better results than other equations. As a result, the AP-FIG and DKIPSO-SVR combined model has extensive application value in the prognosis of stock market opening price.