CAJ | 학술논문

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(FuzzySimplified 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，取得了良好的实验结果.
근거대시간서렬수거진행취류분석적수요，제출료일개가이묘술시량간형태상사도적모호대속도함수표체식，실험표명해표체식제구유교호적시량형태표체능력외，대정구형분포적양본공간야유교호적표체능력.통과대FSART(FuzzySimplified Adaptive Resonance Theory)산법작진일보적개진，제고료타적계산효솔，병장신적대속도함수표체식여개진후적FSART산법상결합，실현료대비평은시간서렬수거적취류분석，취득료량호적실험결과.
For the need of the clustering analysis of time series, a newfuzzy 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.