利用FSART算法实现对时间序列数据的聚类分析

APPLYING THE ALGORITHM FSART TO THE CLUSTERING ANALYSIS OF TIME SERIES

根据对时间序列数据进行聚类分析的需要 ,提出了一个可以描述矢量间形态相似度的模糊隶属度函数表达式 ,实验表明该表达式除具有较好的矢量形态表达能力外 ,对呈球形分布的样本空间也有较好的表达能力 .通过对FSART(Fuzzy Simplified Adaptive Resonance Theory)算法作进一步的改进 ,提高了它的计算效率 ,并将新的隶属度函数表达式与改进后的 FSART算法相结合 ,实现了对非平稳时间序列数据的聚类分析 。

For the need of the clustering analysis of time series, a new fuzzy membership expression that can discribe the modality similarity of vectors is proposed, the results of experiment show that the expression is suitable to the global space too. Some improvement is made to the algorithm FSART to make it more efficient. The new fuzzy membership expression and the improved FSART algorithm are combined to implement the clustering analysis of the nonstationary time series.