A Comprehensive Review of Text Mining Approaches for Predicting Human Behavior using Deep Learning Method

Abstract

This article presents a systematic review of research on predicting human behavior through unstructured textual data, employing a comprehensive selection process illustrated in a flow diagram. The review categorizes 82 selected papers into three primary behavioral domains: emotional, social, and cognitive. Each paper undergoes meticulous examination, identifying objectives, algorithms, computational models, and applications. Natural language processing (NLP) emerges as a dominant text mining approach, utilized in over half of the literature, followed by data extraction, report arrangement, and clusterization. The study further employs VOSviewer to visualize the co-occurrence of the term "text mining," revealing prevalent associations and emphasizing the challenges in analyzing unstructured data efficiently. The article contributes to understanding the evolving landscape of behavior analysis through text mining, addressing the need for automated methods in evaluating individuals' attitudes, emotions, or performance.