An Ensemble Algorithm for Crop Yield Prediction
M Keerthana, K Meghana, Siginamsetty Pravallika, M. Kavitha
Abstract
Machine learning is a pivotal viewpoint for grasping real-world and purposeful use cases for yield prediction of crops. Machine learning is a supportive tool for the agricultural sector which helps us to decide which plant to grow and when to grow the desired plant. This research scrutinizes the usage and implementation of predicting the crop type based on location parameters using ensemble techniques. From a set of given parameters, machine learning can forecast the outcome through unsupervised of supervised learning techniques. To get the required output parameter, we should produce an acceptable and satisfactory function by some set of variables which will depict the output (aimed variable) using the given input variables or parameters. This includes the ensemble (combination) of two machine learning algorithms which improves the crop yield prediction accuracy. Through our searching strategy, we retrieved almost 7 features from various databases and finalized 28242 instances. We investigated these features, analyzed algorithms, and provided propositions for further research work. According to our search strategy, the important parameters taken into consideration are related to climatic conditions like temperature, rainfall, and crop type. From many base papers we came to know that Neural Networks, Decision Tree are the most used algorithms for these models. Decision tree uses parameters like maximum depth and n-estimators, so that by adjusting those parameters, we can get better results. After research, we have concluded that ensemble of Decision tree regressor and AdaBoost regressor gave major accuracy. Crop yield prediction subsumes prediction of the yield of the crop from formerly data. Ultimately, this strategy gives us a recommendation of which crop should be cultivated based on the weather conditions of the field location.