Research on CPI Prediction Based on Space-Time Model

Abstract

Consumer price index (CPI) prediction is an effective approach to measure inflation and provide a reference to formulate economic development strategy. The Autoregressive Integrated Moving Average (ARIMA) model is a classic model to predict CPI. However, a main drawback of ARIMA model is that it only utilizes the time effect while ignoring inter-regional economic interaction which is another significant effect on CPI. Aiming at this, the Generalized Space Time Autoregressive Integrated (GSTARI) model is proposed. In this paper, we verify and compare the prediction accuracy of both GSTARI model and classic ARIMA model with the CPI data of 4 main cities (Dalian, Shenyang, Changchun and Harbin) in China. Our experiments show that for most of cities, GSTARI model have 7%-38% higher prediction accuracy than ARIMA model.