Multi-Step Community Evolution Prediction Methods Via Marcov Chain And Classifier Chain

As social network analysis becomes a popular topic, more and more people show interest in the network community evolution in the future. Currently existing algorithms usually use machine learning methods to predict the future event that a community will undergo. However, these algorithms can't achieve multi-step prediction. This paper proposes two novel algorithms named MC-based prediction and CC-based multi-step prediction. The first one is carried out based on Marcov Chain (MC). In addition to performing multi-step prediction, the MC algorithm is able to achieve higher single-step prediction accuracy than the algorithms using machine learning methods in some cases. The second algorithm proposed in this paper applies the thought of Classifier Chain (CC) to multi-step prediction. Experiments show that these two algorithms are valid and have good performance for multi-step prediction.