A Systematic Review on the Use of Machine Learning Algorithms for Soil Fertility Prediction

Soil fertility assessment is crucial for sustainable agriculture, directly impacting crop productivity and efficient resource management. Traditional assessment methods, while accurate, are often labor-intensive and time-consuming. This study provides a systematic review of Machine Learning (ML) applications in soil fertility prediction, identifying key algorithms, evaluation metrics, and research gaps while also exploring bio-inspired ML models, such as swarm intelligence and genetic algorithms, to enhance predictive accuracy and adaptability. A systematic literature review was conducted on 70 academic papers published between 2012 and 2023, sourced from Google Scholar and Scopus. The study analyzes frequently used ML models, data sources, and performance indicators such as the coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE), and accuracy. The most commonly applied ML algorithms were Random Forest (17.85%), Support Vector Machine (16.13%), and Artificial Neural Networks (5.8%). ANN demonstrated the highest accuracy, with 70% of cases achieving 95%-100% precision, while RF performed well in 75% of cases within the 90%-95% range. Hybrid models showed a broader performance distribution, indicating potential robustness to outliers. The increasing adoption of ML in soil science underscores its potential to revolutionize soil fertility prediction. The results suggest that ML techniques, particularly ANN and RF, provide accurate and efficient alternatives to traditional methods, while hybrid and bio-inspired models offer further improvements. Future research should focus on standardizing ML methodologies and validating models in real-world agricultural settings to enhance their practical implementation.