Communication Theoretic Prediction on Time Series Data
Yunfan Yang, Wen Chen · 2018
With the explosive growth of data, data analytics has emerged as a critical technology and it has attracted considerable research interests. Effective and powerful data analytics are need to deal with big data problems. Network information theory and communication theory have provided some novel methodologies for data analytics in some related works, which presents their feasibility and potential for data analytics in several fresh views. In this paper, an approach based on communication technology is developed for time series data analysis. First, a information channel is established to describe the information flow between two time series data with correlated relationship. Then the structure of communication equalizer is adapted for time series data prediction by inputting with correlated data and history target data. The prediction result shows that the adapted equalizer approach has effectiveness for time series data prediction. For more insights, we modify the update method of tap coefficients inspired by the adaptive filter algorithms. The final experiment results suggest that our proposed approach and its modification have better root-mean-square error performance compared with other approaches in Taiwan Semiconductor Manufacturing Corp stock price time series data prediction, which further demonstrates that network information theory and communication theory have effectiveness and potential and utility for data analytics.