A Survey on Explainable IoT Framework for Soil Fertility Prediction using LightGBM and CatBoost

Abstract

Soil fertility has a direct impact on crop growth and overall agricultural productivity. Understanding soil conditions helps farmers make better decisions regarding irrigation, fertilizer usage, and crop selection. However, traditional soil testing methods mainly depend on manual sampling and laboratory analysis, which are time-consuming and not suitable for frequent monitoring. In recent years, IoT-based solutions have made it possible to continuously observe soil conditions using sensors. These sensors capture parameters such as pH, moisture, and temperature in real time. Along with this, machine learning techniques are used to analyze the collected data and identify patterns that help in predicting soil fertility more effectively. In this survey, attention is given to boosting models such as LightGBM and CatBoost, which are known for handling structured data efficiently. The study also highlights the role of explainable AI techniques like SHAP and LIME, which help in understanding how predictions are generated. Various recent approaches are reviewed, and their strengths and limitations are discussed. Overall, the work focuses on developing systems that are accurate, interpretable, and suitable for real-time agricultural applications.

Introduction

The text discusses the importance of soil fertility prediction in agriculture and how modern technologies like IoT, machine learning, and explainable AI can improve its accuracy and usability.

Soil fertility depends on factors such as nutrients, moisture, temperature, and pH, but traditional soil testing methods are slow, costly, and not suitable for real-time monitoring. To overcome these limitations, IoT-based sensor systems are introduced to collect continuous soil data directly from fields, enabling faster and more efficient decision-making.

Machine learning models like Random Forest, SVM, LightGBM, and CatBoost are widely used to predict soil fertility, often achieving high accuracy. However, many of these models lack transparency, which reduces user trust. To address this, explainable AI techniques such as SHAP and LIME are used to make predictions more interpretable by showing how different features influence results.

The literature survey highlights several existing systems that combine IoT and AI for soil analysis, crop prediction, and nutrient monitoring. While these systems improve accuracy and automation, common limitations include high cost, computational complexity, lack of real-time scalability, and poor interpretability.

The proposed methodology focuses on using IoT sensors to collect real-time soil data (such as pH, moisture, and temperature), which is then processed for further analysis.

Conclusion

This survey presents a sensor-based and explainable approach for soil fertility prediction by integrating IoT, machine learning, and interpretability techniques. The system utilizes real-time soil parameters such as pH, moisture, and temperature to generate meaningful insights without relying on traditional laboratory methods. Machine learning models like LightGBM and CatBoost are used to capture relationships between soil attributes and fertility levels, enabling accurate predictions. Explainable AI techniques such as SHAP and LIME further en-hance transparency by showing how each parameter influences the results. The proposed framework improves the speed and accessibil-ity of soil analysis while supporting better decision-making in agriculture. By combining prediction with interpretability, the system provides a practical and scalable solution for sustainable farming applications.

References

[1] Uwadia O. A., Dahunsi F. M. (2026): IoT and Machine Learning-Based Soil Nutrient Prediction System [2] Priyanshu Tiwari et al. (2026): Real-Time Soil Fertility Monitoring Using IoT and Deep Learning Techniques [3] Kanimozhi Gunasekaran et al. (2025): Soil Fertility Analysis Using Machine Learning and Deep Learning Models [4] M. D. S. Sharafat et al. (2025): IoT-Enabled AI-Based Crop Prediction System with Explainability [5] Remya Praveen et al. (2025): Explainable AI-Based Soil Fertility Prediction Using XGBoost [6] Walid Abdullah (2025): Comparative Analysis of Machine Learning Algorithms for Soil Fertility Prediction [7] C. P. Thamil Selvi et al. (2025): Spatiotemporal Graph Neural Network for Soil Prediction [8] A. Sri Lakshmi et al. (2024): Performance Comparison of LightGBM and CatBoost for Soil Fertility Prediction

Copyright

Copyright © 2026 Sheela Rani C M , Srilakshmi Seshadri, Prabhudeva B N, Shreya Praveen Ramadurg, Sharath M R . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.