Embedded AI: Machine Learning Methods for Superior Plant Recommendations in Microcontrollers
Naufal Widad Sundawa, Muhammad Fahriza Bahrudin, Istiqomah, Khilda Afifah · 2024
By carefully selecting crops depending on soil conditions, agricultural failures can be prevented, and maximum output can be attained. Traditionally, farmers have depended on tradition and imprecise projections, which often leads to disappointing outcomes. Using critical soil parameters like pH, temperature, humidity, rainfall, nitrogen (N), phosphorus (P), and potassium (K), this work closes the gap by recommending appropriate crops using a machine learning (ML) method. We looked at three alternative models—Gaussian Naive Bayes, Random Forest, and XGBoost—to see which machine learning algorithm would be most suitable for incorporating into microcontrollers. As demonstrated by our results, including all seven parameters significantly increases prediction accuracy to 99% when compared to The most effective method, predicting with 100% accuracy in 1.4773 milliseconds, was found when Random Forest was applied to microcontrollers. This work demonstrates the efficient integration of innovative machine learning models into agricultural operations, giving farmers an effective tool to enhance crop selection and increase agricultural productivity. Future research will focus on real-time data collection using sensors to further test and improve the system. using individual or restricted parameter combinations.