关键词: 链接预测, 时间序列, AKIMA模型

Abstract: Existing static link prediction methods have mostly adopted overlay network to represent social network and used structural information of inter-link to predict future link occurrences. However, these methods can only predict new link occurrences, the repeated old link occurrences are not generally studied, so do not apply to many application domains that the prediction of the repeated link occurrences are of main interest. For these deficiencies of static link prediction, this paper introduces the time series link prediction and combines static graph and time series prediction to obtain hybrid time series link prediction. Using the Enron email data the experiments confirm that the time series link prediction can achieve better prediction performance than static link prediction. Furthermore, the hybrid link prediction can get better performance than only using static or time series link prediction.

Key words: link prediction, time series, autoregressive integrated moving average (ARIMA) model