Latent Dirichlet Allocation Feature Extraction with Bio-Inspired Pigeon Feature Selection Technique for Twitter Sentiment Analysis

Abstract

Feature selection act an important role in structure of machine learning. Nonlinear features in the data upset the accuracy performance of the model and rise the training period required to form the system. Feature selection (FS) is asignificant process for making sentiment analysis (SA). In this study, we propose a wrapper FS algorithm for SA. This algorithm uses the technique ofpigeon inspired optimizer (PIO) used to purpose of selection. These algorithm was assessed using three standard datasets: the STC-Twitter dataset. The proposed algorithm overcomes many FS algorithms from the most advanced related conceptssuch as, TPR, accuracy, FPR and F-score. In addition, the projected cosine likeness system for binarizing the algorithm has a quicker integration than the sigmoid scheme. Stimulation outcome display that the proposed PIOFS using the Latent Dirichlet Allocation (LDA) feature extraction technique yields better results.