CAJ | 학술논문

股票市场是个非线性系统，由于受到多方面因素的影响，对于股指的预测一直是个难题。各种建模方法都有自身的缺点，如模式匹配识别系统过分依赖历史数据，缺乏自身变化。 BP神经网络容易陷入局部最优，而且训练时间较长。文章从模式匹配识别和BP神经网络相结合的角度来进行股票指数预测分析，预测系统克服了单一神经网络预测系统和单一模式匹配识别预测系统的各自缺点，能有效地预测股指。
고표시장시개비선성계통，유우수도다방면인소적영향，대우고지적예측일직시개난제。각충건모방법도유자신적결점，여모식필배식별계통과분의뢰역사수거，결핍자신변화。 BP신경망락용역함입국부최우，이차훈련시간교장。문장종모식필배식별화BP신경망락상결합적각도래진행고표지수예측분석，예측계통극복료단일신경망락예측계통화단일모식필배식별예측계통적각자결점，능유효지예측고지。
Stock market is a nonlinear system. Because of its instability to some degree,it’s difficult to predict the stock-index. Some models have their own disadvantage,such as pattern modeling and recogni-tion system rely on the historic data excessively,lacking its own variability. Neural networks is easy to get stuck local optimum and the training time is too long. Based on pattern matching and neural networks,a hybrid system for forecasting stock-index is proposed in this paper. This system can overcome the disadvan-tage of single neural networks system and single pattern modeling and recognition system,and efficiently predict the stock-index.