A Combined Time Series Model for the Prediction of Social Network Popularity and Content Evolution

Abstract

The randomness, dynamism and uncertainty of social networks pose a challenge for revealing the mechanism of content popularity growth. Evolution of content popularity is characterized by strong heterogeneity. It is difficult for a single time series prediction model to capture all kinds of content popularity dynamic evolution patterns at the same time. Using single model to predict complex social network content popularity will lead to poor prediction ability and limited application scenarios. This paper attempts to establish a combined predicting model that integrates the predicting capabilities of multiple traditional time series models. By applying multi-class regression and analyzing the historical prediction performance of each sub-model, the combined weights of the predicted values of each sub-model are generated. The model can learn to adjust the combination weights according to the real-time prediction performance of each sub-model, so as to adapt to the dynamic changes of the evolution model and to enhance the performance of predicting content popularity. The evaluation results of real social network datasets show that the performance of the proposed model is better than that of the existing single models.