关键词: 网络热点话题, 回声状态网络, 相空间重构, 时间序列

Abstract: In order to improve the prediction precision of network hot topic and overcome the defect of traditional of echo state network, this paper proposes a novel hot topic prediction model based on improved echo state network. Firstly, one-dimension time series of network hot topic are reconstructed to multi-dimension, and then multi-dimension time series of network hot topic are input into improved echo state network to establish network hotspot topic prediction model, finally, the performance of model is tested by simulation experiments. The results show that the proposed model can accurately describe the hot topic change trend and improve the prediction precision of network hotspot topic and has good application values.

Key words: network hot topic, echo state network, phase space reconstruction, time series