From Traditional to Advanced Models: A Comparative Study between Time Series and Machine Leaning Models in Agriculture
Praveenkumar A.
ICAR-Indian Agricultural Statistics Research Institute, New Delhi, 110012, India and The Graduate School, ICAR-Indian Agricultural Research Institute, New Delhi, 110012, India.
Manoj Varma *
ICAR-Indian Agricultural Statistics Research Institute, New Delhi, 110012, India and The Graduate School, ICAR-Indian Agricultural Research Institute, New Delhi, 110012, India.
Srinatha T. N.
Division of Agricultural Economics, ICAR-Indian Agricultural Research Institute, New Delhi, 110012, India.
Satyam Verma
ICAR-Indian Agricultural Statistics Research Institute, New Delhi, 110012, India and The Graduate School, ICAR-Indian Agricultural Research Institute, New Delhi, 110012, India.
Naveen G. P.
Department of Mathematics and Statistics, CCSHAU, Hissar, Haryana, 125004, India.
Ankit Kumar Singh
ICAR-Indian Agricultural Statistics Research Institute, New Delhi, 110012, India and The Graduate School, ICAR-Indian Agricultural Research Institute, New Delhi, 110012, India.
Anita Sarkar
ICAR-Indian Agricultural Statistics Research Institute, New Delhi, 110012, India and The Graduate School, ICAR-Indian Agricultural Research Institute, New Delhi, 110012, India.
*Author to whom correspondence should be addressed.
Abstract
The agricultural sector plays a crucial role in the global economy, with edible oil crops like groundnut being vital commodities. Accurate price forecasting is essential for stakeholders, including farmers, traders, and policymakers. The primary aim of this study is to evaluate and compare the effectiveness of traditional time series models (such as ARIMA) and advanced deep learning models (such as RNN, GRU, and LSTM) in forecasting the monthly wholesale prices of groundnut. The analysis covers data from January 2014 to December 2023, collected from Agmarknet. Our results reveal that deep learning models, particularly LSTM, excel in capturing intricate patterns and delivering precise forecasts compared to traditional models. The LSTM model demonstrates superior performance, with RMSE, MAE, and MAPE values of 1.76, 1.02, and 0.25, respectively. This research enhances academic understanding of time series forecasting in agricultural economics and provides valuable insights for refining market predictions and improving decision-making processes.
Keywords: Price forecasting, ARIMA, deep learning, long-short term memory