Feature Analysis and Optimisation for Computational Personality Recognition

Automatically classifying human personality traits through analysis of their social network behaviors is an important yet challenging task to date considering the low accuracy of current researches. In that detection of significant features is an essential part of a personality recognition system, this paper proposes an in-depth analysis of features that contributes to the recognition of a given trait. Besides the common features of social network used by most current researches, text style features and TF-IDF-based psychological features are proposed and prove to be effective to predict certain personality trait. Also particle swarm optimization (PSO) feature optimization algorithm has been adopted to select the best combination of features. Simulation results show that with the best combination of features, the F-measure value of the personality recognition has been improved around 12%.