Work

A Comparative Analysis of Machine Learning Models for Soil Health Prediction and Crop Selection

Year: 2023
Type: article
Abstract: This research paper explores the concept of soil health intelligence and crop recommendation using image analysis techniques. The proposed work focuses on predicting the soil type, pH, suitable crops that can be produced in that soil, and nutrients present in the soil. A soil health intelligence system is proposed in this work by combining machine and deep learning algorithms with image processing techniques. The system is trained with a dataset of soil images and another dataset containing the RGB (Red, Green, and Blue) values along with the pH of an image. The proposed model's ability to predict soil type and properties was assessed through the use of a test dataset, and the findings suggest that the system's accuracy is high, with minimal error. These results highlight the potential of image analysis as a real-world and competent approach for determining the properties of soil in both agriculture and soil science. The image dataset was trained with Convolutional Neural Networks to predict soil type, while the pH-recognition dataset was trained with many regression models, of which the XGBoost Regressor performed the best. Potential benefits of the system include giving farmers, agronomists, and researcher’s vital information on soil management and crop productivity so they may make informed decisions. The findings have major implications for enhancing soil health and promoting sustainable agriculture.
Language: English
Subfield: Plant Science
Open Access status: diamond
OpenAlex ID: W7217021978
FWCI: 5.751
Cites: 36
Cited by: 47
Related to: 10