Early Prediction of Crop Yield in India using Machine Learning
Ankita Sharma, Anushtha Tamrakar, Sourajita Dewasi, Nenavath Srinivas Naik
Abstract
The backbone of India's economy is Agriculture. There is an increased requirement to predict the future crop yield to match the crop demands. Farmers want to know which crop to plant and approximate yield in advance. However, unpredictable rainfall trends, seasonal production trends, and multiple climatic aspects make it challenging to recommend crops and predict yield. Machine learning techniques can resolve the issue. In this paper, we have approached the problem using two models. One of the parts of the model focuses on predicting the crop yield in advance by analyzing factors like district, season, geoclimatic conditions, soil, and crop type. It will help the farmers and the government make agricultural risk management and pricing decisions to max-imize profit. In the models, data pre-processing includes eliminating null values, feature selection and elimination, choosing independent and dependent variables, encoding the categorical variables, and finally splitting the dataset. Random Forest Regressor and Decision Tree Regressor are used for prediction and the metrics used were Accuracy, R2, Adjusted R2 and Residual Standard Deviation. Naive Bayes Classifier, Decision Tree Classifier, KNN Classifier, Random Forest Classifier, Gradient Boosting and XG Boosting were used for the crop suggestion model and the performance metrics used were Accuracy, Precision, Recall and F1 Score. Finally, Random Forest Regressor is considered for crop prediction with an accuracy of 89 % and Random Forest Classifier for crop suggestion with the accuracy of 98%.