Ensemble-based predictive model for crop recommendation

Abstract

Agriculture is a vital industry that supplies food, textiles, and other basic goods to people globally. Agricultural crop production has a vital role in influencing the economy and the well-being of farmers. Nevertheless, farmers are facing substantial challenges due to the profound changes in environmental conditions. A significant challenge they have is determining the most suitable crop for their specific location that will optimize both production and profitability. Choosing suitable crop types for a certain area may be difficult due to the need for skills and experience in evaluating elements such as soil composition, climatic conditions, moisture levels, precipitation, and temperature. Multiple researchers have devised several approaches to tackle the issue of crop recommendation. Nevertheless, a significant share of these models is specifically tailored for a certain job or are amalgamations that include two or three machine-learning algorithms. These current models have restricted prediction accuracy and elevated rates of false positives, rendering them inappropriate for the intricacy of the job at hand. This study explores the field of precision agriculture with the objective of improving crop recommendation systems via the use of an ensemble-based prediction model. This paper incorporates KNN, Decision Tree, Random Forest, SVM, Naive Bayes, Logistic Regression, and XGBoost as a series of machine learning models. A stacked ensemble prediction model is created by training, evaluating, and comparing the Random Forest classifier with the stacked ensemble prediction model. In contrast to existing methods, the proposed method exhibits exceptionally high accuracy, reaching 99.8%, exceeding the performance of prior studies. Through the application of advanced predictive modeling techniques, this paper demonstrates how agricultural operations can be improved.