Performance evaluation of Box-Jenkins' Arima, Box=Tiao's Arimax, and Multiple Linear Regression Models for Rainfall forecasting in Cabanatuan, Nueva Ecija
Date
2023
Authors
Jhoana A. Melendrez
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Rainfall forecasting has gained increasing study interest in recent years because of its substantial impact on agriculture, public safety, and climate monitoring studies. Localized rainfall forecasting models are crucial as the Philippines is projected to have an increased number of heavy daily rainfall and extreme rainfall events by 2020 and 2050. Hence, this study aimed to develop Box-Jenkins', Autoregressive Integrated Moving Avera (ARIMA). Box-Tiao's Auto regressive Integrated Moving Average with Exogenous Variables (ARIMAX), and Multiple Linear Regression (MLR) models to evaluate its performances and select the best rainfall forecasting model for Cabanatuan, Nueva Ecija. In the modelling process, the ARIMA model solely used accumulated rainfall data, the ARIMAX model was incorporated with external variables (temperature, relative humidity), wind direction and wind speed), and the MLR model used predictor variables with a significant relationship to the response variable. The modelling procedures produced the ARIMA (2,0,2) (1,0,0)12, ARIMAX (0,0,0) (2,1,0)12 and y = 1.9737x1-73.7562x2-89.9057x3 (MLR) models. The accuracy of these models was evaluated using Symmetric Mean Absolute Percentage Error (SMAPE), Mean Absolute Error (MAE). Root Mean Square Error (RMSE). Normalized Root M Square Error (NRMSE), and Coefficient of Determination (R2). Result show that the ARIMAX (0,0,0) (2,1,0)12 with exogenous variables of relative humidity, wind direction (cos and sin), wind speed, and temperature is the best-fitted model for monthly accumulated rainfall forecasting in Cabanatuan.