Personality prediction system using machine learning approaches: a comparative study

Abstract

Identifying personality traits from text offers valuable insights for human resource management, customer service, political campaigning, healthcare, fraud detection, and risk assessment. In psychology, personality prediction from text is an important area with the Big Five model, among the leading frameworks. Popular datasets for this task include Essays. Past works have primarily relied on conventional machine learning (ML) models using linguistic features. This paper evaluates and contrasts ten ML classifiers ‘effectiveness for personality prediction using the Essays dataset. The support vector machine (SVM) classifier yielded the best overall performance, a mean accuracy of 57.68% and means F1-score of 61.16% across all five personality traits, and outperformed logistic regression (LR). Thereby demonstrating its superior predictive capability for this task and significance as an interpretable baseline for future deep learning integration.