Abstract
Feature selection is an important research topic in machine learning and pattern recognition. In recent years, data has become increasingly larger in both number of instances and number of features. In fact the number of features that can be contained in a Big Data is hard to deal with. Unfortunately, the number of features that can be processed by most classification algorithms is considerably less. As a result, it is important to develop techniques for selecting features from very large data sets. However the efficiency of existing feature selection algorithms significantly downgrades, if not totally inapplicable, when data size exceeds hundreds of gigabytes. Traditional methods like Filters, Wrappers and Embedded methods lack enough scalability to cope with datasets of millions of instances and extract successful results in a finite time. Therefore, the main purpose of this paper is to propose a new parallel feature selection framework that enable the use of feature selection methods in large datasets.