Mixture Kernel based Least Square Support Vector Machine for Personality Prediction on Social Media

The prediction of the human’s activities and the personality from the social media platforms attains significant growth between the researchers. The statistical data over the human activities represented through the social media platforms is important benefit for the analysis in the predictions of the personalities and other attitudes of the humans. However, the challenges in personality prediction are to introduce effective approach which determines an individual’s personality traits from various behavioural data. Hence, to tackle these issues, this research proposes the Mixture Kernel based Least Square Support Vector Machine (MK-LSSVM) approach for the personality prediction. The utilization of mixture kernels in SVM, enables for modelling the challenging relationships in the data, capturing various types of input features and enhancing the prediction accuracy. The standard datasets such as Essay and Myers-Briggs Type Indicator (MBTI) are utilized to implement the MK-LSSVM approach. The Term Frequency and Inverse Gravity Moment (TF-IGM) approach is utilized for the feature extraction, which determines the relationship between the feature sets and features from the datasets. The implemented results demonstrates that the MK-LSSVM approach reaches the superior average accuracy of 89.72% and average precision of 0.89 as compared to the CatBoost algorithm.