摘要
介绍了利用支持向量机与重构相空间理论预测混沌时间序列的方法,并以股价时间序列为样本,比较了几种常用核函数的预测能力,实验表明高斯核的预测能力明显好于其它核.使用遗传算法优化了高斯核支持向量机的参数,优化后其预测能力较经验定参方法有明显提高,且好于传统的预测方法.
The Support Vector Machine Theory is a hotspot in the field of machine learning in recent years. In this article a chaotic time series prediction method using Support Vector Machines and phase construction theory is introduced, and the predicting performances of several common kernel functions are compared, taking stock price time series as samples. Experiments show that Gaussian kernel obviously performs better than other kernels. Support Vector Machines with Gaussian kernel are optimized by Genetic Algorithm, and performs much better than SVMs whose parameters are decided by experience, and also better than traditional prediction methods.
基金
国家自然科学基金(70171053)资助项目.
关键词
预测
支持向量机
混沌时间序列
遗传算法
核
predication
support vector machine
chaotic time series
genetic algorithm
kernel