Forecasting Comparisons of PAR, DLM, ETS, and ARIMA Time Series Models with Weather Data

• Activity Number: 264
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Yingyu Tang* and Jong-Min Kim
• Companies: University of Minnesota, Morris and University of Minnesota, Morris
• Keywords: Time series analysis ; Weather Forecasting

Abstract

We compared four statistical time series methods with weather data. The first method we used is Periodic Autoregressive Model, the second method is Dynamic Linear Model, the third method is Exponential Smoothing State Space Model and the fourth method is Autoregressive Integrated Moving Average Model. With these four models, we compared the accuracy of forecasting with ME (Mean Error), MPE(Mean Percentage Error) MAE(Mean Absolute Error), MAPE (Mean Absolute Percentage Error) and MASE (Mean Absolute Scaled Error). The final conclusion is that Autoregressive Integrated Moving Average Model and Exponential Smoothing State Space Model performed the better accuracy of forecasting compared with Dynamic Linear Model and Periodic Autoregressive Model.