Weekly Price Forecasting of Tomato in Malaysia Using Sarima: A Short-Term Seasonal Analysis

Abstract

Tomato price fluctuations can influence agricultural planning, stability and decision making in the market between the producers and consumers. The proper forecasting models thus become significant in forecasting the future price changes. This study aims to forecast the weekly tomato price in Malaysia with the use of Seasonal Auto-Regressive Integrated Moving Average (SARIMA) model under Box-Jenkins approach. A total of 208 observations of the weekly tomato price data of 2021 until 2024 were analyzed using RStudio software. The modelling process involved exploratory analysis, stationarity testing, model identification, estimation, validation and application forecasting. The Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) were used to compare candidate SARIMA models. It was found that the SARIMA(3,1,1)(1,0,0) model had the lowest value of the AIC and BIC and was considered the final model. The residual diagnostic results, including the Shapiro-Wilk and Ljung-Box test, confirmed that the residuals satisfied the model assumptions, indicating that the selected model was adequate for forecasting. The forecasted results showed that tomato prices are expected to fluctuate over the next 52 weeks.