A Novel Time Streams Prediction Approach Based on Exponential Smoothing

A data stream prediction algorithm using Linear Regression based on Exponential Smoothing method was proposed in this paper, namely Exponential Smoothing based Linear Regression Analysis (ES_LRA) data stream prediction algorithm. The ES_LRA algorithm only processes the current sliding window, which can improve the operation efficiency; In the meantime, it applied a Smoothing Coefficient(α) through the exponential smoothing method to smooth the estimate parameter in order to eliminate the concussion caused by unit data and increase the prediction accuracy. The experiment simulated a building fire by FDS4.0 and estimated the trend of the fume temperature. Analytical and experimental evidence show that the ES_LRA algorithm performs better both on prediction accuracy and operate efficiency.