Py_FS: A Python Package for Feature Selection Using Meta-Heuristic Optimization Algorithms

Abstract

Guha, Ritam Chatterjee, Bitanu Khalid Hassan, S. K. Ahmed, Shameem Bhattacharyya, Trinav Sarkar, RamIn today’s data-driven world, every workforce is relentlessly exploiting the power of data to get that extra edge in order to triumph over the others. However, there is a saying that goes like, “Work smarter, not harder.” Studies have shown that the amount of data people actually use is way smaller than the data being generated. This very fact gives rise to an important research topic, called dimension reduction, which is one of the smartest (not hardest) strategies for retrieving useful information (here, features) from a given high-dimensional dataset. Feature selection (FS) is among the best dimension reduction tactics available in the literature. To this end, in this paper, we have made an effort to introduce an FS framework called Py_FS that we have developed to simplify the task for the researchers. Py_FS currently provides an interesting combination of 12 wrapper- and 4 filter-based FS techniques along with various evaluation metrics. For converting the continuous search space to a binary search space, three transfer functions have been used. The algorithms have been experimented on two Microarray and four UCI datasets. To the best of our knowledge, it is the first ever framework to provide wrappers, filters, and evaluation metrics under one structure. The framework is highly flexible and can easily cater to the needs of the FS researchers. It is publicly hosted at the following link: Py_FS: A Python Framework for FS .