蔡孟哲

For an enterprise, it is fundamental to win as many customers as possible. The key to successfully winning customers is often determined by understanding the personality characteristics of communication objects in order to employ an effective communications strategy. An enterprise needs to obtain the personality information of target or potential customers. However, the traditional method for personality evaluation is extremely costly because of time and labor consumption, and it is incapable of acquiring customer personality information without their awareness. Therefore, the manner in which to effectively conduct automated personality predicts for a large number of objects is an important issue. The diverse social media that have emerged in recent years have become a digital platform where users deliver their speeches publicly and interact with others. Perhaps social media can serve the needs of automated personality predicts. Based on Facebook user data, the main social media platform in the world, this research developed three methods for predicting personality types based on interactions logs and users’ statuses. In this research, Dominance, Inducement, Submission, Compliance (DISC) proposed by Marston is used as the personality model. To predict personality types, the interaction features of users were designed accordingly and calculated, and some text mining technique such as TF-IDF, VSM, and normalized edit distance were used in this research. For interactions, this research also serves to build a universal model for social media interaction, and it is used to propose an efficient method for designing interaction features; for users’s statuses, this research developed a complete mechanism based on crowdsourcing, and it could make real Facebook users provide their data and label training data efficiently.