A deep learning and multi-objective PSO with GWO based feature selection approach for text classification
Pradip Dhal, Chandrashekhar Azad · 2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE) · 2022
Text Classification (TC) is becoming an increasingly important activity due to the vast quantity of documents available on the internet, through emails, and in digital libraries. It's usually accomplished after completing Feature Selection (FS), which entails choosing relevant characteristics to improve classification accuracy. The cost of computing is reduced, and the accuracy of the TC system is improved by diminishing the dimension of feature space. As a result, one of the essential tasks in TC is to detect the optimal feature set. In this work, we have introduced a hybrid framework that is based upon Multi-Objective Optimization (MOO)-based FS and for the classification, we have used a Bidirectional Long Short Term Memory (BiLSTM) network for the TC system. Here we have used a modified version of Particle Swarm Optimization (PSO) with Grey Wolf Optimization (GWO) for the FS task. In this multi-objective FS, the first objective function is to enhance the Classification Accuracy (CA), and the second one is to diminish the selected features. The experimental findings suggest that the proposed hybrid model accomplish competitive performance compared with the other traditional classifier.