Imbalance data classification algorithm based on SVM and clustering function
Kai-Biao Lin, Wei Hsiu Weng, Kuo-Hua Robert Lai, Ping Lü · 2014
The traditional support vector machine (SVM) was mainly used well on balanced data classification, but didn't perform well at imbalance dataset classification. In order to improve classification effects of SVM algorithm for imbalance dataset, the present paper combined the merits of FCM cluster algorithm and SVM algorithm to create a new algorithm (referred as FCM-SVM algorithm). Meanwhile, we adopted F-measure evaluation indicators, combining with predicting accuracy and recall of minority class, to evaluate algorithm classification performance. Effectiveness of FCM-SCM algorithm was verified by repeated experiences on dataset from UCI Database, the result shows that the algorithm improved the classification performance for imbalance problem compared to existing SVM algorithms.