Use of Machine Learning Technique in Forecasting the Price and Arrival of Onion in Important Markets of Anugul District of Odisha, India
Devidatta Behera
Department of Agricultural Statistics, College of Agriculture, Bhubaneswar, Odisha University of Agriculture and Technology, India.
Abhiram Dash *
Department of Agricultural Statistics, College of Agriculture, Bhubaneswar, Odisha University of Agriculture and Technology, India.
*Author to whom correspondence should be addressed.
Abstract
Price fluctuations in onion crop of Odisha, is due to variation in production and market arrival. Study of the market arrival and market price of onion in important and efficient markets of Anugul district of Odisha is conducted in the study. The study can help policymakers create effective agricultural policies to reduce price volatility and provide timely information to the farmers to adopt best cropping pattern and sell their product in market at appropriate time and obtain adequate profit.
Data pertaining to arrival and price of onion crop in market of Anugul district has been collected for the period from April 2022 to March 2024. The traditional ARIMA model can take care of only linearity in the data. Since data on price and arrivals of crops are usually nonlinear in nature, the modern technique of models i.e Machine Learning models are fitted to the data on arrival and prices of onion. These modern models would fit the data having non-linearity nature. ANN models with different nodes at hidden layer are fitted to the data. The best fit model selected by appropriate model diagnostic tests and model fit such as, Root Mean Square Error (RMSE) and Mean Absolute Percent Error (MAPE) are cross validated and finally used for prediction purpose.
Neural Network Auto Regressive models at different nodes and hidden layers are fitted separately to the data for arrival and for price in Anugual, Athamalik and Jarapada market of Anugul district are found to be the best fit models for forecasting. The forecasted values of onion price of Anugul and Athmalik markets are found to be stable and in Jarapada, market it is found to be increasing and then stable with time. The forecasted values of arrivals of onion are found to be stable in Jarapada market. In Anugul market arrivals of onion is found to be decreasing and then stable while in the Athmalik market the forecasted values of arrivals of onion are found to be fluctuating with alternative increase and decrease.
Keywords: ANN, fluctuation, MAPE, model diagnostics test, RMSE