International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Agriculture is a vital sector that supports the livelihood of millions of people, particularly in developing countries like India, where farming remains a primary source of income. However, the sector faces persistent challenges such as improper crop selection, soil nutrient imbalances, climate variability, water scarcity, and lack of scientific decision-making. Traditional farming practices rely heavily on farmers’ intuition, experience, or generalized recommendations, which often result in poor crop yield, excessive fertilizer usage, and increased vulnerability to environmental fluctuations. To address these issues and support precision agriculture, this research proposes a machine-learning-based Crop Recommendation System that utilizes soil nutrient values—Nitrogen (N), Phosphorus (P), and Potassium (K)—along with environmental factors such as temperature, humidity, pH level, and rainfall to recommend the most suitable crop for cultivation in a given region.

The proposed system uses supervised machine learning algorithms, including Random Forest, Decision Tree, Naive Bayes, and Support Vector Machine (SVM), to analyze large agricultural datasets and learn complex patterns between soil–climate features and crop suitability. The dataset undergoes extensive preprocessing, which includes handling missing values, normalizing numeric attributes, removing noise and outliers, and encoding categorical labels. Feature engineering further enhances the prediction quality by identifying the most influential variables such as the N:P:K ratio, soil fertility index, and climate–soil interactions. These features help the model better distinguish crop requirements under varying environmental conditions.

During the experimental phase, each algorithm was trained and evaluated using standard performance metrics, including accuracy, precision, recall, and F1-score. Results show that the Random Forest classifier outperformed other models, achieving an accuracy of 97%, largely due to its ensemble nature, robustness, and ability to handle high-dimensional data. The findings highlight that machine learning can significantly improve agricultural decision-making by offering farmers scientific guidance tailored to their land conditions. This system has the potential to enhance crop productivity, minimize risks associated with crop failure, optimize fertilizer usage, and promote long-term soil health.

This research demonstrates that integrating machine learning into agriculture provides a practical and scalable solution to modern farming challenges. Future advancements may include IoT-enabled soil sensors, satellite-based remote sensing, real-time data analysis, and deep learning models for yield prediction and dynamic crop recommendation. Overall, the study emphasizes the transformative potential of machine learning in supporting sustainable and smart agriculture. This research aims to develop and evaluate a machine-learning-based crop recommendation system using key soil and climatic features. By comparing multiple ML algorithms and identifying the most accurate model, the study contributes to the growing field of smart agriculture and demonstrates how technology can transform traditional farming practices.

Downloads: | DOI: 10.17148/IJARCCE.2025.141181