Portable Field Device for Real-Time Soil Fertility Prediction Using CatBoost

Accurate prediction of soil fertility is the main factor to improve agriculture production with adapted sustainable farming. Therefore, this paper would present a proposition for developing a Soil Fertility Prediction System using the NPK-pH sensors integrated with Arduino to acquire the data in real time. The system will incorporate the use of the CatBoost model, a gradient boosting algorithm, in predicting soil fertility using sensor data. The Arduino-based sensor setup allows for constant monitoring of soil conditions, providing a very accurate and highly precise measure of NPK levels and pH values. These are very key indicators of soil health that may influence crop yield and quality. Data collected is fed into the CatBoost model that is trained on historical soil and crop yield datasets to understand the relationship between soil properties and fertility. Experimental results reveal quite high performance of the CatBoost model in predicting soil fertility, therefore showing good application of the model. This system will be beneficial in making appropriate decisions for fertilization and good management of the soil by farmers, leading to better yields of crops and usage of resources efficiently. This was a huge step forward in precision agriculture, where sensor technology was integrated with high-machine-learning models, showing the potential of IoT-based sensor networks along with machine learning for real-time assessment of soil fertility and sustainable agricultural practices.