基于自适应遗传算法的股票预测模型研究

Stock prediction model research based on improved adaptive genetic algorithm

为了解决单一神经网络模型很难满足股票预测建模要求的问题,提出一种基于遗传算法的粗糙集属性约简方法和神经网络相结合的预测模型。在该模型中,改进了自适应性遗传算法的交叉算子与变异算子。基于该遗传算法的粗糙集属性约简相比传统的粗糙集属性约简,其具有更强的求解最小属性约简的能力,解决了神经网络预测时训练速度慢、内存开销大等问题;在数据预处理过程中,引入聚类分析,有效解决了连续属性离散化的问题。实验结果证明,该预测模型具有较高的预测精度,在时间序列的股票预测中是相当有效的。

It is difficult for a single neural network model to meet the modeling requirements of stock prediction. To solve the problem, a prediction model which combines the neural network with the rough set attribute reduction method based on the genetic algorithm is proposed. The model improves the crossover and mutation operator of the adaptive genetic algorithm. Comparing with the traditional one, the rough set attribute reduction based on the adaptive genetic algorithm is more capable to obtain the minimum attribute reduction, which solves the drawbacks of the neural network prediction model such as slowly training speed, large memory overhead and so on;In the data preprocessing phase, the utilization of cluster analysis could effectively solve the problem of discretization of continuous attributes. Experimental results prove that this predicting model is more accurate and fairly effective in the time-series stock prediction.