Ensemble ARIMA-LSTM Algorithm in Predicting Agriculture Commodities Market Price for Farmer’s Education

Agriculture is crucial as the backbone of national economy and source of food for human. To be adapted with Industry 4.0, farmers should possessed the knowledge on price prediction of the agriculture crops and commodities so that they can make better farming decision according to supply and demand. In this paper, we analysed the agriculture commodities market price data of three categories namely fruit, poultry and vegetable. Six commodities of berangan banana, chicken, choy sum, green chilli, highlands tomato and standard chicken were chosen for analysis based on their data availability and most commonly used in Malaysian dishes. It was observed that the vegetable category is the most volatile in terms of price fluctuations. Green chilli had the highest variation while berangan banana showed the least. Three artificial intelligence based algorithms namely Autoregressive Integrated Moving Average (ARIMA), Long Short-Term Memory (LSTM) and ensemble ARIMA-LSTM algorithms were employed for price prediction using 12 datasets with a total of 7235 daily price data at retail level. The results show that ensemble ARIMA-LSTM performed better in predicting most of the commodities as determined by RMSE values. The proposed algorithm should be versatile in predicting market price at any demographic, in any country in the world.