A new modification and application of item response theory‐based feature selection for different machine learning tasks

Feature selection (FS) is an important step of the existing machine learning (ML) methodology since it often makes it possible to obtain better results using a lower number of features. Hence, in the literature, there exist many studies aiming at proposing a new FS method or improving an existing one in the context of ML. Accordingly, this study presents a new lossy modification of a feature selector which is a specific type of filter-based FS and depends on item response theory. This method computes feature importance in a supervised manner and is previously employed for classical text categorization (TC) task, where it was shown that the selector provided satisfying results on high dimensional and benchmark text datasets. As such, this paper introduces a new modification of this selector along with its new variants and investigates its applicability for different ML tasks other than TC. Experimental results are obtained on 35 different datasets, of which nine are well-known and real-world datasets from the UCI ML repository. Our comparative results with the most popular filter-based FS methods show that it is possible to obtain better results with this new modified selector or one of its variants on the majority of both binary and real-world datasets compared to its well-known peers.