A comparison of different weather forecasting models for the monthly forecast of Lahore city

Abstract

In this paper, we study the performance of different statistical models used in weather forecasting and compare their forecast accuracy. In particular, we use time series regression (TSR), seasonal autoregressive fractional integrated moving average (SARFIMA) and artificial neural network (ANN). A dynamic non-linear autoregressive (NAR) back-propagation ANN algorithm is also applied to estimate the forecasting accuracy. For ANN model, we use the moving average (MA) and Holt-Winter exponential smoothing (HW-ES) transformations for pre-processing the data. The monthly data of different weather parameters are obtained from the Pakistan Meteorological department to apply the aforementioned models. The results show that the ANN model with the MA transformation of the data has the smallest root mean squared error and the highest correlation coefficient for different weather parameters.