Modern Implementations of Feature Selection Algorithms and Their Perspectives

Abstract

In the last decade, due to the growing interest in neural networks and machine learning in general, the Python programming language became the main language for many data scientists and machine learning engineers. This rapid growth resulted into the lack of many key machine learning algorithms in the existing Python libraries. Feature selection, as one of the main fields of data preprocessing for high-dimensional, was not covered properly in Python language, although it can be widely used to improve models quality and solve some of the overfitting problems. In this paper we have performed a review of existing open-source Python feature selection libraries, made their comparison and pointed their drawbacks and after that presented our own open-source ITMO FS library. Moreover, we have added some examples on its usage and compatibility with most popular modern machine learning Python library scikit-learn and some performance tests.

Keywords

• machine learning
• dimensionality reducion
• feature selection
• python
• programming libraries