Crop Yield Prediction to Achieve Precision Agriculture using Machine Learning

Crop yield prediction is one of the most demanding tasks in agriculture. It is crucial in making decisions at global, regional, and field levels. Numerous factors such as genotype, environment, and their interactions decide a mosaic trait such as crop yield prediction. Faultless yield prediction demands elementary understanding of the relationship between yield and the associated factors. Comprehensive datasets and powerful algorithms are required to reveal the connection between the associated factors. Machine learning is an indispensable tool for crop yield prediction, which helps in making decisions with regard to the type of crops to be grown during the season and what to be done during the growing season of crops. An accurate crop prediction model helps farmers in deciding the financial investment and forecast the returns on the investment. The proposed solution aims to create a heaped model consisting of linear models. In order to achieve the solution, data from the period of 1995-2010 were collected, coming from official varietal experiments carried out in India. The solution consists of a random forest regressor, decision tree regressor, KNN regressor and XGB regressor. For validation of the model, four forecast error metrics were used: i.e., mean absolute error (MAE), mean squared error (MSE), R2-SCORE and Accuracy. As a result of the conducted experiments, the models individually provided an accuracy ranging from 63% to 85% and the stacked model provided an accuracy of 94%.