Purpose–Due to the unintentional or even the intentional mistakes arising from a survey, the purpose ofthis paper is to present a data-driven method for detecting students’friendship network based on their dailybehaviour data. Based on the detected friendship network, this paper further aims to explore how theconsidered network effects (i.e. friend numbers (FNs), structural holes (SHs) and friendship homophily)influence students’GPA ranking.
Design/methodology/approach–The authors collected the campus smart card data of 8,917 sophomoresregistered in one Chinese university during one academic year, uncovered the inner relationship between thedaily behaviour data with the friendship to infer the friendship network among students, and further adoptedthe ordered probit regression model to test the relationship between network effects with GPA rankings bycontrolling several influencing variables.
Findings–The data-driven approach of detecting friendship network is demonstrated to be useful and theempirical analysis illustrates that the relationship between GPA ranking andFNpresents an inverted“U-shape”, richness inSHspositively affects GPA ranking, and making more friends within the samedepartment will benefit promoting GPA ranking.
Originality/value–The proposed approach can be regarded as a new information technology for detectingfriendship network from the real behaviour data, which is potential to be widely used in many scopes.Moreover, the findings from the designed empirical analysis also shed light on how to improve GPA rankingsfrom the angle of network effect and further guide how many friends should be made in order to achieve thehighest GPA level, which contributes to the existing literature.