Author(s): Riya Thakur, Subhash Sharma and Anmol Negi

Abstract: Apple is an important fruit crop of Himachal Pradesh, accounting for almost 49% of the total area under fruit crops and 85% of total fruit production. Fluctuations in the prices of agricultural crops affect supply and demand and have a significant impact on consumers. Accurate prediction of agricultural commodity prices would facilitate the reduction of risk caused by price fluctuations and is of great significance to the farmer’s economies. Various time series models viz. ARIMA, ARCH-GARCH, and Recurrent neural network long short-term memory (RNN-LSTM) were used for efficient price prediction. The Solan market was selected purposively based on the highest arrival of apple produce in the state. To evaluate, these models daily price data was collected from AGMARKNET for the year 2012 to 2023. In all models, the best-fitted model was selected based on minimum information criteria. The results showed that using ARIMA (6, 1, and 1) and GARCH (1, 1) models were the best-fitted models. However, to confirm the validity of the models, the Root Mean Square Error value (RMSE) and Mean Absolute Percentage Error (MAPE) were compared which shows that the RNN (LSTM) model performed comparatively well over other models for forecasting apple prices. The prediction results based on the RNN model were better than those of the separate ARIMA and GARCH models. Furthermore, it best fits the actual price profile and has better generalizability.

Pages: 237-243 | Views: 182 | Downloads: 6

Download Full Article: Click Here