An Iterative Algorithm of Key Feature Selection for Multi-class Classification
Daeun Jung, Hyunggon Park · 2019
In this paper, we propose an iterative algorithm of key feature selection for multi-class classification problems, where the data includes a large number of features but the amount of data is limited. For efficient classification, the proposed algorithm first extracts a set of key feature candidates based on Boruta algorithm and then iteratively adopts conventional machine learning based classification algorithms to determine key features. Simulation results show that the proposed algorithm can effectively determine key features, leading to improved classification accuracy compared to direct adoption of multi-class classification algorithms.