Density-Ratio Peak Based Semi-Supervised Algorithm for Access Network User Behavior Analysis

In order to improve the prediction accuracy of the access network user behavior (ANUB), we propose a novel density-ratio peak (DRP)-based semi-supervised algorithm. It first rescales the given dataset with non-uniform density clusters by density-ratio estimation (DRE) and conducts the subscriber detailed classification by using the density peak (DP) algorithm. The proposed DRP algorithm can identify all clusters in a dataset with greatly varying densities. Then, a semi-supervised algorithm evolves three regression prediction methods, namely, an auto regressive and moving average (ARMA), an auto regressive integrated moving average (ARIMA), and a fractionally auto regressive integrated moving average (FARIMA), as typical representatives to generate an accurate predictor for the ANUB and establish a prediction model for each subcluster. The behaviors of access network users in the same subcategory share similarities and the statistics of the behaviors of all access network users in a district can emerge the model of the district network properties, such as the prediction of the district network traffic that is more detailed than the direct prediction. The proposed model is evaluated through the dataset of the ANUB collected from China Telecom, and the obtained results show that the integrated model is an effective way to improve the accuracy of prediction achieved by the DRP clustering, compared with the conventional ones.

Abstract licence: CC BY-NC-ND

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.