在這篇論文中,我們提出了一個新的替代式模糊短時間序列聚類演算法。本論目的在於觀察此新演算法是否比模糊短時間序列聚類演算法 [3]更加具有優勢。在使用模糊短時間序列聚類演算法 [3]因為其建立在模糊C均值演算法之上,所以對於具有離群值的短時間序列資料作聚類分析時有些缺點,使得分類結果不正確。所以為了改善此問題,在此篇論文中我們將Moller-Levet教授 [3]提出的Short Time Series距離與吳國隆教授及楊敏生教授 [4]提出的替代式模糊C均值演算法結合為替代式模糊短時間序列聚類演算法,此演算法建立在替代式模糊C均值 [4]之上,因此對於具有離群值之時間序列資料做聚類分析時可以改善分類結果,且可以改善模糊短時間序列演算法的缺點。
In this thesis, we propose Alternative Fuzzy Short Times Series Clustering Algorithm (A-FSTS). This thesis was to observe whether the new algorithm is better than Fuzzy Short Time Series Clustering Algorithm [3]. We using Fuzzy Short Time Series Clustering Algorithm (FSTS) [3] for Short Time Series data with outliers have some problem and the classification results are incorrect, because Fuzzy Short Time Series Clustering Algorithm [3] is based on Fuzzy C-Means Algorithm (FCM). We want to improve this problem, So we will combine distance of Short Time Series [3] and Alternative Fuzzy C-Means (AFCM) [4] as the Alternative Fuzzy Short Time Series Clustering Algorithm. This algorithm based on the Alternative Fuzzy C-Mean Algorithm, Therefore, this algorithm is not affected outliers, and improve the classification results.