您现在的位置首页 » 科学研究 » 科研成果 » 正文

To minimize the negative effects of the outliers in the training data set during the data description modeling, a support vector data description algorithm(SVDD) based on fast clustering analysis is proposed. The proposed approach consists of two stages of strategy. Firstly, a fast clustering analysis algorithm is applied to preprocessing the training data set. The outliers in the training dataset that affect the model are removed. Then, a weighted support vector data description based on kNN is used in the second stage. Experimental results on benchmark datasets show that the performance of the present approach is superior to conventional SVDD and density-induced SVDD in accuracy.

版权所有: 同济大学CIMS研究中心 地址:上海市曹安公路4800号同济大学嘉定校区电信学院大楼(智信楼)

电话:021-65988911-8628 传真:021-65983673  邮箱:cims@tongji.edu.cn