Prediction of Personality Trait using Machine Learning on Online Texts

Personality is an amalgamation of a varied set of features and characteristics of an individual that effects their state of mind, attitude towards others, day to day activities and line of thoughts and reasoning. All the text of an individual that is available on social media sites provide us an opportunity to recognize personality traits of that individual. XGBoost is a machine learning method that uses gradient boosting framework. In this work that is proposed we use it foretell four personality traits that are associated with Myers-Briggs type indicator MBTI model and they are judging perceiving, intuition sensing, feeling thinking, introversion, and extroversion. All the experiments are conducted on readily accessible dataset from Kaggle. The inaccuracy or the less accuracy of the previous data are mainly credited to the skewness of the dataset. This can be changed by using resampling method on this dataset. This method is also termed as over sampling which results in better accuracy or the overall performance of the mechanism which we employ here. For more exploration of the personality from text, pre-processing techniques including tokenization, word stemming, stop words elimination and feature selection using TF IDF are also exploited. All the pre-processing methods including word stemming, stop word removal, feature selection, tokenization is applied to further improve the exploration of the personality. All these works would help the for the improvement of personality identification system and these in turn would help and assist different organisation in their employment of appropriate people in their organisation. This also enables them to improve their business by understanding the needs and requirements of their customers. All the results that are acquired from classifiers are good enough, but the performance delivered by the XGBoost classifier is a whole new level and its enables us to achieve 99% precision for various traits.