Comparative Study of Feature Selection Algorithms for Text Analytics: A Performance and Efficiency Analysis

In text analytics, high-dimensional datasets frequently impede model performance, rendering feature selection an essential process for enhancing accuracy and computing efficiency. This study examines and contrasts various feature selection algorithms, classified into filter, wrapper, and embedded techniques, within the realm of text analytics. The study evaluates the effectiveness of approaches including ChiSquared, Recursive Feature Elimination (RFE), and L1 Regularization on benchmark text datasets to ascertain their competence in dimensionality reduction while maintaining model performance. The findings illustrate the trade-offs between computational cost and performance, offering insights into the optimal use of feature selection techniques for various text classification tasks.