Novel machine learning technique for analyzing personalities in status updates on facebook

Abstract

The largest and widely used online social network, Facebook keeps extensive records of user activity as it is expressed in a variety of activities, including Facebook likes, status updates, posts, comments, photographs, tags, and shares. A social media user's tweets and other actions make their psychological foundations quite evident. However, making a prediction about this is difficult. This study investigates how to predict the HEXACO Model Personality Scores, which offer a quantitative assessment of users' personality characteristics, using integrated random forest fused extreme gradient boosting (IRF-XGB). Assemble a database of Facebook status updates and the corresponding personality labels. Natural Language Processing (NLP) methods and the Bag of Words (BoW) tool were used to preprocess the obtained data and extract features. In terms of predicting HEXACO model personality traits, the proposed IRF-XGB yields better results. The prediction accuracy results show that the personality prediction method developed with the IRF-XGB classifier is higher than the average baseline for all the feature sets, with a top prediction accuracy of 97%, even when tested using the identical dataset.