Feature Selection Based on Graph Representation

Best features subset identification is an important preprocessing step in Machine Learning and Data Mining. Therefore, many feature selection algorithms have been proposed in the literature. Generally, there are three major approaches of feature selection: Filters, Wrappers and Embedded. In this paper, we propose a new feature selection approach for numerical datasets, which is based on graph representation where the node degree used as criterion to select the best subset of features among the whole features space. The experimental results show the effectiveness of the proposed algorithm in terms of execution time and achieved performance.