模糊核聚类支持向量机集成模型及应用

Support vector machine ensemble model based on KFCM and its application

为了进一步提高支持向量机在回归预测中的精度,提出一种基于模糊核聚类的最小二乘支持向量机集成方法。该方法采用模糊核聚类算法根据相互独立训练出的多个LS-SVM在验证集上的输出对其进行分类,并计算每一类中的所有个体在独立验证集上的泛化误差,然后取其中平均泛化误差最小的个体作为这一类的代表,最后经简单平均法得到集成的最终预测输出。在短期电力负荷预测中的实验结果表明,该方法具有更高的精确度。

To further enhance the regression prediction accuracy of support vector machine, a Least Squares Support Vector Machine （LS-SVM） ensemble model based on Kernel Fuzzy C-Means clustering （KFCM） was proposed. The KFCM algorithm was used to classify LS-SVMs trained independently by its output on validate samples, the generalization errors of LS-SVMs in each category to the validate set were calculated of the LS-SVM whose error was minimum would be selected as the representative of its category, and then the final prediction was obtained by simple average of the predictions of the component LS-SVM. The experiments in short-term load forecasting show the proposed approach has higher accuracy.