Enhancing Crop Yield Prediction Through Advanced Data Mining Techniques

Abstract

Crop yield prediction aids resource allocation and agricultural decision-making. Integration of several data sources and effective data preprocessing and feature selection are needed for accurate prediction. Machine learning procedures improve model performance and interpretability via normalization and feature selection. This paper has proposed crop yield prediction using ensemble-based normalization and feature selection methods with SVM Classification. The ensemble normalization has utilized with Average filling, Weighted K-means clustering and Decision tree algorithms. Weighted K-means clustering and decision tree assigns values to samples based on their distances from cluster centers to show data distribution. An average filling fills missing values with the average of their properties, completing the dataset for analysis. Next, the feature selection has utilized Random Forest (RF), Logistic Regression (LR), PCA and Elastic Net selects important features. Principal components analysis optimizes representation and feature selection by selecting orthogonal components that best reflect data variation. The last step is classification using Support Vector Machine (SVM). The SVM model has classify the new instances using the important features. To improve crop yield production using rainfall, humidity, N, P, K and pH attributes are considered. These factors are crucial to crop health and growth. The SVM classification has achieved 91% accuracy while using the ensemble normalization and feature selection methods used.