Classification of university students' behaviors in sharing information on Facebook

Online social networks, particularly Facebook has become one of the most popular platforms for students to make connections, share information, and interact with each other. In modern socializing society, many distinct patterns can be deduced from the observations. These specific traits provide direct benefit to businesses as students can sometimes act as spokesman for their merchandises on social media without extra investment. In this paper, the implications of Facebook “share” with respect to commercial gain are analyzed based on students' behaviors. An interaction matrix of “share” interaction and profile data are composed as a dataset which are clustered into six eligible groups of commercial segment: dining, itinerary, pets, entertainment, games, and gifts/varieties. Pervasive classification algorithms: KNN, Decision Tree, NaïveBayes and SVM are applied to explore the opportunity of target products. According to our findings, SVM outperforms the others with accuracy of 87.95 % due to its distinctive characteristic in handling imbalanced data. The classification results reveal that the merchandises that have high potential in the campus are entertainment CDs, itinerary and pets. This valuable result can also be expediently applied to new-coming students.