Prediction Of Myers-Briggs Type Indicator Personality Using Long Short-Term Memory

Abstract

Personality is defined as the mix of features and qualities that make up an individual's particular character, including thoughts, feelings, and behaviors. With the rapid development of technology, personality computing is becoming a popular research field by providing users with personalization. Many researchers have used social media data to automatically predict personality. This research uses a public dataset from Kaggle, namely the Myers-Briggs Personality Type Dataset. The purpose of this study is to predict the accuracy and F1-score values so that the performance for predicting and classifying Myers–Briggs Type Indicator (MBTI) personality can work optimally by using attributes from the MBTI dataset, namely posts and types. Predictive accuracy analysis was carried out using the Long Short-Term Memory (LSTM) algorithm with random oversampling technique with the Imblearn library for MBTI personality type prediction and comparing the performance of the method proposed in this study with other popular machine learning algorithms. Experiments show that the LSTM model using the RMSprop optimizer and learning speed of 10^-3 provides higher performance in terms of accuracy while for the F1-score the LSTM model using the RMSprop Optimizer and learning speed of 10^-2 gives a higher value than the proposed machine learning algorithm so that the model MBTI dataset using LSTM with random oversampling can help in identifying the MBTI personality type.