Development of a Real-Time Soil NPK Prediction System Using AS7265X Spectral Sensor and Random Forest Regression

VIS-NIR spectroscopy technology is used to measure soil NPK content through spectral reflectance, providing an accurate solution for real-time soil monitoring. However, the main challenge is the high dimension of spectral data that is difficult to analyze, and traditional models such as linear regression are less effective in handling non-linear relationships. This study developed an NPK prediction system using the AS7265X sensor and the Random Forest algorithm, which is expected to improve the accuracy of Nitrogen, Phosphorus, and Potassium prediction. The results showed that the Random Forest model was able to predict NPK with very good accuracy ($\mathrm{R}^{{2}} {0. 9 9 9 6}$ for Nitrogen, ${0. 9 9 0}$ for Phosphorus, and 0.970 for Potassium). This indicates that Random Forest is more effective in handling non-linear data and is suitable for field applications.