A Novel Model-Based Approach in Classification Using Extension Distance
In order to solve the low efficiency problem of KNN or K-Means like algorithms in classification, a novel extension distance of interval is proposed to measure the similarity between testing data and the class domain. The method constructs representatives for data points in shorter time than traditional methods which replace original dataset to serve as the basis of classification. Virtually, the construction of the model containing representatives makes classification faster. Experimental results from two benchmark data sets, verify the effectiveness and applicability of the proposed work. The model based method using extension distance can effectively build data models to represent whole training data, and thus a high cost of classifying new instances problem is solved.