基于最小二乘支持向量机的模糊时序分析方法

Fuzzy time series analysis method based on least squares support vector machines

利用最小二乘支持向量机良好的分类和函数估计能力,提出了一种新的模糊时序分析方法.该方法包括两部分:在模糊时序处理部分通过建立启发式规则、模糊变量、论域、模糊集合和隶属度函数,完成历史数据的模糊化;最小二乘支持向量机处理部分替代传统的模糊关系计算,对模糊化的历史数据进行分析,然后去模糊化得到最后的预测结果.与多种传统模糊时序分析方法的对比试验表明,该方法充分利用了支持向量机较好的推广性能等优点,具有更高的精度和较好的泛化效果.

Aiming at the problem of low precision of traditional fuzzy time series （FTS） analysis methods, this work proposed a new method based on least squares support vector machines （LS-SVM）, which is an efficient tool for pattern recognition and regression estimation. This method includes two parts. In the FTS processing part, establish heuristic rules and fuzzy variables and determine the universe of discourse, fuzzy sets and membership degree functions, then fuzzily the history data. The LS-SVM processing part uses LS SVM instead of traditional fuzzy relationship computation to analyze fuzzy data, and produces the final results after defuzzifieation. Comparing with traditional FTS analysis methods, this new method can obtain higher accuracy and good generalization quality.