0
- Volume 2, Issue 1 2024
By Hira Farmaan, Noman Hassany, Noman Islam
10.20547/aibd.242105
Keywords: Time Series, Rainfall, Vector Auto Regression (VAR), and ARIMA (Autoregressive Integrated Moving Average)
The challenge of predicting rainfall is undoubtedly a difficult one because there are an extensive number of different factors and elements that influence the conditions of the climate. It is of extreme significance to have accurate rainfall forecasts, particularly for the agricultural sector, which is highly reliant on timely and sufficient rainfall for the growth and yield of crops. The contribution that agriculture makes to the economy is another factor that highlights the need of accurate rainfall forecasts. There are many different ways that have been utilized all around the world in order to forecast the patterns of rainfall. These methods include classical statistical methods that are frequently used to advanced machine learning techniques. Thus, in this work, two quite opposite methods of creating a model are described. Such approaches comprise of statistical method which employs the Autoregressive Integrated Moving Average (ARIMA) alongside with the vector auto regression (VAR) model. The Box-Jenkins method provided the foundation on which the ARIMA and VAR modelling procedures employed in this study were built on. These techniques consisted of four primary stages: fitting or estimating a model, making predictions, assessing the model’s quality, and evaluating the model’s parameters. The purpose of these stages was to assess the performance of the prediction process and for this a mean annual rainfall data for each year from 1901 to 2016 of country Pakistan was used.. The models were trained using data spanning 116 years, which included the mean annual rainfall. When applied to the data, Time series forecasting methodologies are utilized in order to produce the optimization coefficients and the regression coefficients, correspondingly. Based on the findings of the study. The ARIMA model doesn’t fit the data as well as the VAR model does. This is evident from the fact that the RMSE, MAE, MAPE, AIC/BIC, POCID, and R² numbers are smaller and stronger, respectively. In this case, the VAR model is better at predicting how much rain will fall each month and throughout the year. The government, researchers, and private individuals should take into consideration these univariate and multivariate models when making plans for the future in order to increase agricultural commodity production and prevent damages caused by excessive precipitation.