OPTIMIZING AUTOMATED PERSONALITY INSIGHT: A HYBRID MACHINE LEARNING APPROACH

Abstract

This scholarly article explores enhancing automated personality prediction using combined machine learning methodologies. Utilizing the Myers-Briggs Type Indicator (MBTI) dataset sourced from Kaggle is central to this investigation, specifically focusing on refining the accuracy of predicting the Judging-Perceiving (J/P) dichotomy. Past research has underscored notable hurdles in forecasting this aspect, frequently resulting from challenges such as data leakage and model inefficiencies. In order to confront these barriers, a range of machine learning algorithms, specifically Logistic Regression, Random Forest, K-Nearest Neighbors (KNN), and XGBoost are examined in this analysis. The findings indicate that ensemble models, especially XGBoost, perform superior to traditional single-model approaches in terms of precision and reliability.

Additionally, advanced data preprocessing methods like TF-IDF vectorization and SMOTE are integrated to address data imbalance issues and further elevate model efficacy. By amalgamating these diverse modelling techniques, this research establishes a sturdy framework for automated personality forecasting, supplying valuable insights for fields such as psychology, marketing, and human resources. By bridging critical gaps in existing methodologies and proposing innovative remedies for more precise and dependable personality prediction, this paper makes a significant contribution to the academic domain.