Integrating Decision Tree and KNN Hybrid Algorithm approach for Enhancing Agricultural Yield Prediction

Abstract

Accurate prediction of agricultural yield plays a pivotal role in ensuring sustainable resource allocation and global food security. Traditional methods often struggle to capture the intricate relationships between diverse agricultural variables, necessitating innovative approaches for enhanced prediction accuracy. This paper presents the Decision Tree-KNN Hybrid Algorithm (DT-KNN), a novel method that integrates decision trees and K-nearest neighbors (KNN) to adopt these challenges effectively. Decision trees are recognized for their aptitude to model complex interactions and interpretability, making them suitable for capturing nonlinear patterns in agricultural data. On the other hand, KNN excels in local pattern recognition by utilizing similarities between data points. By combining these two methodologies, DT-KNN leverages the strengths of both to enhance predictive precision and robustness. The methodology begins with comprehensive data preprocessing, incorporating cleanup, standardization, and attribute production. This stage guarantees that the input data is standardized and optimized for subsequent modeling. The decision tree component of DT-KNN constructs a hierarchical structure that partitions the information into splits based on characteristic estimates, thereby identifying distinct patterns in the agricultural data. Each leaf node of the decision tree represents a subset of data points with similar characteristics. Subsequently, KNN is applied within each identified leaf node to make localized predictions. This dual-layered approach allows DT-KNN to capture both global trends and local variations within the Indian Chamber of Food and Agriculture (ICFA) agricultural dataset, thereby improving the overall predictive accuracy. To validate the effectiveness of DT-KNN, extensive experiments are conducted using ICFA datasets. The performance of DT-KNN is evaluated against traditional methods and other hybrid algorithms through rigorous comparative analysis. System of measurement such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared (R2) are engaged to assess predictive accuracy and robustness across diverse algorithms. The results demonstrate that DT-KNN outperforms traditional methods in terms of accuracy and reliability. It effectively balances between capturing complex agricultural dynamics and maintaining interpretability, making it a promising approach for agricultural yield prediction. This research aids to the improvement of predictive modeling in farming and lays the groundwork for future enhancements and applications of hybrid algorithms in agricultural research and practice.