Predicting Self-Reported Proactive Personality Classification With Weibo Text and Short Answer Text

Abstract

Personality assessments are at present nearly entirely dependent on self-reports, and machine learning methods have been rarely applied to this field. This study used machine learning to predict people's self-reported proactive personalities. Based on a sample of 901 participants that used Weibo text and short answer text, the authors used five machine learning algorithms for classification: Support Vector Machine (SVM), XGboost, k-nearest neighbor (KNN), naïve Bayes, and logistic regression. Seven different indicators - Accuracy (ACC), F1-score(F1), Sensitivity(SEN), Specificity (SPE), Positive Predictive Value (PPV), Negative Predictive Value (NPV) and Area under Curve (AUC) - combined with hierarchical cross-validation were also used to make the comprehensive evaluation of models. Based on this, we proposed a method to classify people's proactive personalities based on text mining technology. The results showed that the SVM and naïve Bayes outperformed the other methods on short texts (i.e., short answer texts) and mixed long texts (i.e., short answer & Weibo text). In the context of the texts used, mixed long text (i.e., short answer & Weibo text) improved and stabilized the indices, and this combination was the best choice of text for predicting proactive personality. In addition, the SVM was the most stable classifier in most situations, even on Weibo text that was not suitable for analysis as long text, and it also recorded good results in terms of accuracy, the F1-score, and the AUC.