ANOVA Validation and Machine Learning Metrics Verification for Crop Yield Prediction Based on Soil Parameters, Climatic and Temperature Conditions

Abstract

India is agricultural state and its economic system usually relied on the agricultural yield growth and agro-industrial merchandise. One of the difficult ventures in an agricultural area is crop yield production. The existing available prediction methods doesn’t take into consideration the rainfall, weather conditions, soil and its characteristics as a whole.  Hence the combined constructional and recommendation system is proposed for crop prediction. The algorithm which includes the Artificial Neural Network (ANN) and demographic model of Multiple Linear Regressions is used as a recommendation system within the agricultural places to increase the yield of crops. The artificial neural network algorithm is less prone to error than other machine learning algorithms and its one of the efficient tool for prediction methods. The ANN back propagation algorithm fine tunes the weights of the neural network based on the error rate obtained from the previous iterations. Fine tuning is done to reduce the error rate. MLR is to model linear relationship between the explanatory (independent) variables and response (dependent) variable. This paper presents a brief analysis of crop yield recommendation system by combining MLR – ANN methods for the entire region of India. A prediction system is developed by combining a MLR - ANN approach that took advantage of state-of-the-art modeling and solution techniques. Our model was found to have a higher prediction results, with a root-mean-square-error (RMSE) being 5.5% of the average yield and 4.4% of the Mean Absolute error value (MAE) for the validation dataset. Moreover, the accuracy rate is increased by 98.9% for the combined approach. The experimental results shows that the proposed work efficiently predicts the crop yield production and the results are verified using the statistical method, Analysis of Variance (ANOVA).