基于信息熵的支持向量回归机训练样本长度选择

Selection of Training Sample Length in Support Vector Regression Based on Information Entropy

支持向量回归机(support vector regression,SVR)是近年来发展起来的一种通用的机器学习方法。目前已被广泛应用于工业、经济等很多领域,取得了良好的效果。但对于大规模非平稳数据的训练学习,会因为规模较大和样本长度选择的问题,影响到预测结果的精度。为了有效缩减训练样本长度,选择出合适的训练样本,提出基于信息熵的训练样本长度选择方法。该方法利用信息熵对数据的平稳性进行度量,从而选择出最平稳的数据进行学习。该方法不但减少了数据长度、节省了学习时间,同时也提高了预测结果的精度。

Support vector regression（SVR） is a general machine learning method in recent years.It has been widely used in industry,economy and many other fields and has achieved good results.However,when the non-stationary data in large quantity to be studied,the accuracy of results will be influenced by the problem of larger number and the selection of training sample length.In order to eliminate the length of training sample and select the proper training sample,a method of selecting of training samples length based on information entropy was proposed.This method evaluates the stationarity of the data by the information entropy and then picks out the most stationary data for learning.According to the experiments,the method not only saves learning time,but also improves the accuracy of predictions.