Predicting Peer Offline Probability in BitTorrent Using Nonlinear Regression

Dongdong Nie, Qinyong Ma, Lizhuang Ma, Wuzheng Tan

Abstract: BitTorrent is a popular and scalable P2P content distribution tool. This study attempts to analyze the factors that affect the offline probability of BitTorrent peer, and express the probability using these factors. We first collect large data set of BitTorrent peers’ activities. Then we use nonlinear least-squares regression to determine the probability distribution function for each of the three factors (download percent, download speed, and local time) and the joint probability distribution function of the three factors, and use another large data set to verify the prediction results.