SMART FARM-CROP RECOMMENDATION SYSTEM USING MACHINE LEARNING

Abstract

As the agricultural sector faces growing challenges such as soil degradation, unpredictable weather patterns, and changing climatic conditions, the need for data-driven decision-making in farming has become essential. Traditional crop selection methods often rely on human intuition or static rules, which can lead to inefficiencies and inaccurate predictions. To address this, the present study introduces a Smart Farm Crop Recommendation System that utilizes multiple machine learning algorithms—Logistic Regression, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Decision Tree—to predict the most suitable plant type based on soil and environmental parameters. The dataset, titled Plant_Parameters_14cols, includes 14 key soil and climatic attributes such as pH, soil moisture, and nutrient composition. After preprocessing steps including encoding, normalization, outlier removal, and data balancing using SMOTE, the models were trained and evaluated through standard performance metrics. Among all tested algorithms, the Random Forest classifier achieved the highest accuracy, precision, and recall, demonstrating its robustness for crop prediction tasks. The findings emphasize that machine learning can effectively enhance precision agriculture and provide intelligent support systems for smart farming. Future work will focus on integrating real-time sensor data and cloud-based applications to improve the system’s adaptability and scalability.