Measuring intimacy levels based on mobile communication patterns

Abstract

In this paper, I suggest a method to measure the intimacy scores of interpersonal relationships using mobile communication data. Referencing studies in the literature, I define the communication factors that affect intimacy as Frequency (the number of communications), Intensity (communication duration), and Time and Day (the time at which communication occurred), Directivity (the person who mostly initiated the communication between user and his/her partner), and Channel (tendency to choose between calls or SMS). Using these factors, I implemented a method of pattern-based intimacy prediction. I used similarities (measurement: Pearson R) between the communication factors for hierarchical clustering (method: furthest neighbor) of users with similar communication patterns. I also used three criteria for clustering communication factors: 1) factor similarities (Pearson R), 2) the number of data in a cluster, and 3) the intimacy standard deviation. The optimal combination of criteria was a high similarity of factors within a cluster, a large amount of data in a cluster, and a low standard deviation for the intimacy scores of a cluster. The mobile communication data of 14 Korean people were used for reference data, and 33 communication patterns were derived. For the evaluation, data of 10 additional Korean people were used. According to the each pattern representative values (matching to the factor centroid values), the additional data were classified to the best similar pattern (Pearson R). After that, the centroid intimacy value of each pattern was provided to the additional classified data. Finally, the correlation of original intimacy to pattern closely matched predicted intimacy was 0.61(p<0.01). Using this method, we can automatically predict interpersonal relationship; the type of interpersonal relationship between the users can be determined through the character of their mobile communications.