Prediction of MBTI Personality Leveraging Machine Learning Algorithms

Abstract

In this study, the author attempted to implement a machine learning approach to determine users' corresponding MBTI personality types by relying only on the content of their online forum postings. Models based on different algorithms are built and trained, and the natural language of the collected data set is converted into machine language for machine learning and used in subsequent tests to determine the correctness of the predicting results. The data set is collected from the forum and divided into two parts, the training set is leveraged to train the model and the test data set is leveraged to make personality predictions and compare with the training data set to measure the correctness of the predicting outcomes. The results show that logistic regression algorithm and vectorized representation of text with TfidfVectorizer can best accomplish the prediction task. This study completed a preliminary comparison of algorithms for personality prediction from text, which became the basis for subsequent personality model predictions using other media.