Data Intensive Parallel Feature Selection Method Study

Feature selection is an important research topic in machine learning and pattern recognition. It is effective in reducing dimensionality, removing irrelevant data, increasing learning accuracy, and improving result comprehensibility. With the development of computer science, data deluge occurs in many application fields. Classical feature selection method is out of work in processing large-scale dataset because of expensive computational cost. This paper mainly concentrates on the study of data intensive parallel feature selection method. The parallel feature selection method is based on MapReduce program model. In each map node, a novel method is used to calculate the mutual information and combinatory contribution degree is used to determine the number of selected features. In each epoch, selected features of all map nodes are collected to a reduce node and from which a feature is selected through synthesiation. The parallel feature selection method is scalable. The efficiency of the method is illustrated through an example analysis.