Investigation into How Machine Learning Can Be Used on Social Media Data to Better Understand Personality During Recruitment Process

Abstract

The traditional recruitment process is time-consuming and incurs additional financial costs. In the digital era, the text shared on social media platforms can be valuable to analyse user’s behaviuor based on user information. People use social media networks to communicate and share true their insights. This provides an opportunity to use that information to detect personality traits. The automatic detection of personality traits can effectively find a suitable applicant for the required job. This study explores the relationship between an individual’s MBTI type and various aspects of their writing style. This study uses a publically available dataset to detect personality traits. Using an MBTI-rated dataset of user posts in the Personal Forum, this study presents two approaches, one approach is based on binary classification, and the other approach is based on multiclass classification, by classifying each individual into one of the 16 types. This research uses three traditional machine learning classifiers (i) Support Vector Machine, (ii) Linear Regression, and (iii) Random Forest using TF-IDF weighting scheme to predict user’s personality types from their posts. This study uses 10-fold cross- validation to evaluate results. Finally, the results reveal that the logistic regression algorithm outperformed other classifiers and obtained 95% average accuracy.