基于AP-SVM组合模型的股票价格预测

Stock Price Trend Prediction Based On AP-SVM Combined Model

为了减小单支股票训练数据中的噪声对分类器性能的影响,提出了一个新的基于簇的股票价格涨跌预测方法(AP-SVM)。AP-SVM首先使用近邻传播(AP)算法挑选出与待预测股票价格变化相似度较高的其他股票,然后将待预测股票和与其价格变化相似的其他股票一起作为输入数据,训练一个支撑向量机(SVM)实现对待预测股票价格涨跌的预测。实验结果表明,当训练数据中存在噪声时,AP-SVM在预测准确率方面优于传统的SVM方法。

To overcome the performance degradation of the classifier due to the noise in individual stock price data,this paper proposes a novel cluster-based stock price trend prediction method(AP-SVM). We first used the affinity propagation algorithm to select all the stocks with high similarity in the aspect of the price tendency to the predicted stock,and then used these stocks as the training data to fit a support vector machine (SVM). The experimental results show that AP-SVM surpasses the traditional SVM method in predicting accuracy in the noisy training data.