Predicting Vegetable Prices in Sri Lanka Using Machine Learning Techniques

Abstract

Sri Lanka is an agriculture-based country, nearly 33.7% of households are farm dependent. As farmers they are contributing to the Sri Lankan economy. The major problem that the farmers face, is when the vegetables are not worth the price and farmers are unaware of the marketing price. Vegetable price prediction is very challenging due to many reasons such as climate changes, demand, supply etc. But, predicting the vegetable prices are essential for the Sri Lankan economy, agriculture sector, farmers as well as consumers to make effective decisions and to prevent the loss of social welfare due to excess supply and excess demand. During the last decade, couple of studies were used traditional statistical techniques like ARIMA to forecast the crop prices in Sri Lanka. However, no study was utilized machine learning techniques like novel based approach to predict the vegetable prices in the Sri Lankan context. This study was presented different machine learning techniques such as gradient boost, XG boost, random forest regression and stack regression techniques which were used to predict the vegetable prices in Sri Lanka. Utilized models for each vegetable were assessed based on the performance matrices and the best performing model for each vegetable was suggested for the future vegetable price prediction. As the source of data, daily price reports of vegetables published in the Central Bank of Sri Lanka from 2016-2022 was used.