A Comparison of the Forecasting Models of Rainfall Data of Umudike, Abia State Nigeria

FULL TEXT in - | page 49-57

Abstract

The forecasting ability of some Time series Models in forecasting the rainfall data of Umudike was investigated. Monthly rainfall data of Umudike in Abia State were collected from Meteorological Department of National Root Crops Research Institute (NRCRI) Umudike from 2008-2017. Triple Exponential Smoothing (Holt-Winters Method), Multiple Linear Regression approach and SARIMA Model were employed to model the data. Mean Error (ME), Mean Square Error (MSE), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) and Friedman Statistic test were used to ascertain the forecasting ability or performance of the models. The Friedman test for the significant difference among the models showed that there was no significant difference among the three models. Further, based on the accuracy measures, it was shown that the Multiple Linear Regression approach has lower accuracy error measures than other models considered, and as such it is regarded as the most appropriate model for forecasting rainfall pattern of Umudike, when judging from the view of Accuracy Measures. It was also deduced from the analysis that Friedman test Statistics is the most appropriate method for testing the performance of competing models, since it makes use of statistical tools for inference.

Keywords: Accuracy Measurement, Forecasting models, Rainfall, Time series,

References