Time Series Data Analysis on Rice Production in Sri Lanka

Abstract

Rice is Sri Lanka's staple meal, eaten almost every day by a significant proportion of its people. Thus, rice production is generally centered on population, price, producers, related industries, and government authorities. There are two main rice production seasons: the Maha season (from September to March) and the Yala season (from May to August). In Sri Lanka, rice is said to have a regal past. More than just a staple dish for this island nation, its significance extends far beyond that. Rice symbolizes the nation and is vital to its history, customs, and even politics. Building a set of models to determine the long-term pattern and forecast future developments in paddy production for leading years is also a secondary objective. Further, rice is a major export crop of Sri Lanka. Therefore, by understanding the pattern and predicting the amount of rice production in the future, it is possible to be prepared in case the amount of rice production decreases. The study used secondary data from the Department of Census and Statistics, Sri Lanka, from 1951 to 2019. Two models were built mainly considering Yala and Maha seasonal rice production data and yearly rice production data, respectively. Autoregressive Integrated Moving Average (ARIMA) (5, 1, 0) was the most suitable model for seasonal data as it has the lowest Akaike information criterion (AIC) and Bayesian Information Criteria (BIC) values with 19.923 Mean Absolute Percentage Error (MAPE) value. Consequently, the 3MA (3- Moving Average) model was developed, which provided a Mean Absolute Error (MAE) of 200.556 and MAPE of 9.13. Considering both results, the most suitable model to forecast rice production in Sri Lanka was determined to be the 3MA model. In light of the findings, the researchers infer that 3MA is the best technique for predicting rice output in the future. Buyers and sellers will be able to use the results to plan for rice production in the coming years, as well as identify periods of low output and investigate their causes.