Comparison between Different Mustard Yield Prediction Models Developed using Various Techniques for Udaipur Region of Rajasthan

Abstract

The present research was conducted to develop and compare mustard yield prediction models using SPSS regression, Artificial Neural Network (ANN) and Autoregressive Moving Average (ARIMA) model. Mustard crop is one of the major rabi crops of India with Rajasthan as the leading mustard producing state. In this study, eight different weather parameters were used to develop mustard yield prediction model, with different yield prediction techniques. Weather and yield data from year 1999 to 2015 were utilised for calibration and year 2016 to 2018 for validation. Three different algorithms were used in ANN to predict mustard yield. Time series model (ARIMA) is another technique used in this study to forecast mustard yield for Udaipur district. In order to analyse and compare error(s) in the developed models and to compare simulated and actual/observed yield, different error indices like root mean square error (RMSE), standardized root mean square error (SRMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and D Index were considered. The validated results showed that, regression model spss performed better than other two models as the RMSE value using SPSS model was very less (0.12 to 0.14), also the D index value using regression model was close to 1.