An Embedded Backward Feature Selection Method for MCLP Classification Algorithm

Feature selection is very crucial for improving classification performance, especially in the case of high-dimensional data classification.Different classification algorithms tend to select different optimal feature subsets. Based on detailed analysis of the characteristics of Multiple Criteria Linear Programming (MCLP) classification algorithm, a feature selection criterion is presented and an embedded backward feature selection procedure is designed for MCLP in this paper. Experiments on four datasets (artificial and real-world) are carried out, and the effectiveness of the presented method is assessed. Results show that it achieves good performance as expected