Research And Realization Of Classification Analysis

As a kind of machine learning algorithm which is based on the statistics learning theoretics, the support vector machine (SVM) has excellent classification performance. Understandable classification model is very important for human experts. However, currently, the classification model for SVM classifiers is non-understandable by human experts. In this paper, we introduce four boolean kernels. Using the four kernel functions which we introduce, we build a novelty decision rule classifier named as DRC-BK. Experiment results show that, for both text classification and non-text classification, DRC-BK has excellent classification performance. It is a disadvantage that DRC-BK will produce a large size of rule set. In this paper, we also present a weighted SVM normal rule selection algorithm, which efficaciously resolves this problem. Experiment results show that, after applied with this algorithm, the size of rule set for DRC-BK is greatly reduced, at the same time, classification performance of DRC-BK is improved a little.