ANN based crop yield prediction from remotely sensed retrieved crop parameters using machine learning
Krupavathi K, M Raghu Babu, A Mani, PRK Parasad and L Edukondalu
The application of remote sensing in crop studies at regional level is becoming very famous. Crop yield prediction and mapping at different scales other than field scale is a challenging task for the researchers using remote sensing. This article described about the successful development of a scientific model using artificial neural network to predict the crop yield on regional scale using well known feed forward and back-propagation algorithms with the help of remotely sensed retrieved crop parameters. The Feed Forward Back Propagating Neural Network (FFBPNN) model developed and was calibrated using the remote sensing retrieved parameters and ground truth data in Mat lab environment. The model gave accurate and stable results. The highest mean relative error was 6.166% and the lowest relative error was 0.133%. To test the performance of the developed model statistically, Coefficient of determination, root mean squared error, mean absolute error and the average ratio of predicted yield to target crop yield (Rratio) and relative error were used. This study also tested the number of hidden neurons on the performance of the model. The statistical analysis confirmed the reliability of the developed ANN model for its applicability on remote sensing-based parameters for paddy yield estimation (The range of R2 values are from 0.933 to 0.992 for training and same for testing it ranged from 0.928 to 0.989). Based on the results, it was concluded that the FFBPNN models performed better and could be applied successfully to estimate and map the crop yield of paddy.
How to cite this article:
Krupavathi K, M Raghu Babu, A Mani, PRK Parasad, L Edukondalu. ANN based crop yield prediction from remotely sensed retrieved crop parameters using machine learning. Pharma Innovation 2021;10(5):1191-1200.