The recent area of interest for computer scientists and data analysts working on precision farming has been the use of machine learning and deep learning algorithms to recommend crops to the farmers and predict yield. Improving crop yields not only helps farmers but also seeks to address global problems such as food shortages. Predictions can be made taking into account forecasts of climate, soil and its mineral content, moisture, crop historic performance, rainfall and others. Crop Data from six states of India for five crops from 2009–2016 has been used for training and validation of different machine learning regression algorithms to produce a comprehensive study of crop yield estimate. The yields are estimated using Linear models, Support Vector Regressor, K Neighbors Regressor, Tree-based models, Ensemble models and Shallow Neural Networks with R-squared score for evaluation. Test accuracy showed promising results in ensemble models and neural networks. Extra Trees Regressor is the best model with Mean Absolute Error of 351.10 and maximum accuracy of 99.95%. This paper aims at providing state of art implementation on machine learning algorithms to facilitate farmers, governments, economists, banks to estimate the crop yields in Indian states based on specific parameters.
原文链接
分享
引用
