A Parameter Based Customized Artificial Neural Network Model for Crop Yield Prediction
K. Aditya Shastry, H. A. Sanjay, Abhijeeth Deshmukh · Journal of Artificial Intelligence · 2015
Background: Selection of the crop for planting is one of the major challenges faced by farmers.Crop selection is influenced by many factors like the weather, nature of soil, market, etc. Weather and soil type are the major factors which affect the crop yield.Crop yield prediction helps the farmers in the selection of the crop for plantation.Crop yield can be accurately predicted by considering the parameters like nature of the soil, amount of rain, crop characteristics, etc. Methodology: There are couple of methods which can be used to predict crop yield.Artificial Neural Networks (ANN) and Multiple Linear Regression (MLR) are two well-known prediction techniques.In this study prediction of the wheat crop yield is done by considering parameters like amount of rainfall, crop biomass, soil evaporation, transpiration, Extractable Soil Water (ESW) and amount of fertilizer applied (NO 3 ).Default-Artificial Neural Networks (D-ANN) is a ANN with only one hidden layer.In this study Customized ANN (C-ANN) is developed by varying the number of hidden layers, number of neurons in the hidden layer and the Learning Rate (LR).Experiments are conducted to compare the C-ANN with MLR and D-ANN models on the same dataset using R 2 statistic and percentage prediction error.Results: Results show that the C-ANN model performs better with a higher R 2 statistic and a lower percentage prediction error than the MLR and D-ANN models on the test dataset.Conclusion: Prediction of crop yield is very important in the community of agriculture.In this study wheat yield was predicted by considering its different parameters.Better wheat yield was predicted by using C-ANN model.