Toll Free Helpline (India): 1800 1234 070

Rest of World: +91-9810852116

Free Publication Certificate

Vol. 11, Special Issue 6 (2022)

Comparison of machine learning and regression approaches to forecasting Alternaria blight epidemic of Indian mustard

Author(s):
Manjari Singh, Parul Setiya, Anand Kumar Tewari and Ajeet Singh Nain
Abstract:
In the present investigation, weather-based prediction models have been developed for predicting epidemic characteristics of Alternaria blight of mustard. Models used in the study are ENET, LASSO, Ridge, and ANN which have been developed using 14 years (2006 to 2020) of epidemics data on (1) crop age at first appearance of Alternaria blight, (2) crop age at highest disease severity, and (3) highest disease severity in a growing season. Models were trained with 70% of data (2006- 2016) and remaining 30% data (2017-2020) were used for testing the model. Performance evaluation was done using R2, root mean square error (RMSE), normalized root mean square (nRMSE), Mean Biased Error (MBE), and modeling efficiency (EF). Results indicate that models performed well at the calibration stage for all variables at all sowing dates. However, at validation stage, ANN derived models gave excellent results (R2val = 1.00, nRMSEV ~ 0.00 to less than 5, and MBEV less than 1 in most cases), and LASSO derived models gave satisfactory results. Evaluation metrics (including R2val, nRMSEV, and MBEV) suggested that ENET- and Ridge-derived models do not perform satisfactorily, whereas ANN-derived models yielded reliable results.
Pages: 2974-2984  |  290 Views  107 Downloads
How to cite this article:
Manjari Singh, Parul Setiya, Anand Kumar Tewari and Ajeet Singh Nain. Comparison of machine learning and regression approaches to forecasting Alternaria blight epidemic of Indian mustard. The Pharma Innovation Journal. 2022; 11(6S): 2974-2984.

Call for book chapter