Forecasting of Vegetable Prices using STL-LSTM Method

Agricultural product prices play an important role in the agricultural market. Vegetables have the largest supply and price fluctuations among agricultural products. As vegetables are grown outdoor and their yields change considerably according to meteorological changes, it is difficult to stabilize the supply and prices of vegetables. Thus, vegetables have a large effect on the national economy. Although the government makes many efforts to stabilize the supply and prices of vegetables, but frequent meteorological changes in recent years have led to unstable supply and price fluctuations of vegetables. Therefore, the correct forecasting of vegetable prices is an important issue. To deal with such an issue, this study suggests a vegetable price forecasting model that uses the seasonal-trend-loess (STL) preprocessing method, and long short-term memory (LSTM), a deep learning algorithm. The model was used to forecast monthly prices of vegetables using vegetable price data, meteorological data of chief producing districts, and other data. In this study, the model was applied to Chinese cabbages and radishes in the Korean agricultural market. The results of performance measurement show that the suggested vegetable price forecasting model had forecast accuracies of 92.06% and 88.74%, respectively, about Chinese cabbages and radishes. It is expected that the model can be used for an autonomous adjustment of supply demand and to develop relevant policies in order to save social costs in relation to agricultural product yields.