TY - JOUR
AU - Vanitha, G.
AU - Kennedy, J. S.
AU - Prabhu, R.
AU - Rajkishore, S. K.
PY - 2021/07/19
Y2 - 2022/08/18
TI - Trained neural network to predict paddy yield for various input parameters in Tamil Nadu, India
JF - Journal of Applied and Natural Science
JA - JANS
VL - 13
IS - SI
SE - Research Articles
DO - 10.31018/jans.v13iSI.2812
UR - https://journals.ansfoundation.org/index.php/jans/article/view/2812
SP - 135 - 141
AB - <p>The major objective of the present study was to explore if Artificial Neural Network (ANN) models with back propagation could efficiently predict the rice yield under various climatic conditions; ground-specific rainfall, ground-specific weather variables and historic yield data. The back propagation algorithm will calculate each expected weight using the error rate as the activity level of a unit was altered. The errors in the model during the training phase were solved during the back-propagation. The paddy yield prediction took various parameters like rainfall, soil moisture, solar radiation, expected carbon, fertilizers, pesticides, and the long-time paddy yield recorded using Artificial Neural Networks. The R<sup>2 </sup>value on the test set was found to be 93% and it showed that the model was able to predict the paddy yield better for the given data set. The ANN model was tested with learning rates of 0.25 and 0.5. The number of hidden layers in the first layer was 50 and in the second hidden layer was 30. From this, the testing value of R square was 0.97. The observations with the ANN Model showed that i) the best result for the test set was R<sup>2 </sup>value of 0.98, ii) the two hidden layers kept with 50 neurons in the first layer and 30 neurons in the second one, iii) the learning rate was of 0.25. With all these configurations, maximum yield is possible from the paddy crop.</p>
ER -