Modeling Analysis Based on Live Broadcast Paid Gifting Behavior

Viewers, on the Internet, can purchase and send virtual gifts to express their love for the streamers. This is not only a novel behavior, but also a blank field of related academic research. To explore paid gifting behavior, this paper uses the data from the DouYu platform to find out the regularity of individuals and groups of viewers. From the individual level, we find that the user's time interval for sending virtual gifts obeys an exponential distribution, random and uniform. In addition, through the T-test to find people with considerable changes in giving virtual gifts, three typical individuals were be analyzed, of which three individuals liked to paid gifting the same streamers, who were primarily true love fans. From the group level, according to the dynamics of human behavior, the number of virtual gifts sent by groups obeys a power-law distribution. We use the Apriori algorithm to study frequent set analysis and the changing law between group types, find people like to watch live game broadcasts, and prefer to give gifts to Beauty, Outdoor and Dance. Finally, we use unsupervised clustering to divide the paid gifting group users into low, medium and high consumption three categories and analyze each type, then found that users will send danmaku when they are sending virtual gifts.