Pre-treatment Spectral Data NIRS Menggunakan Support Vector Regression

Abstract

Potassium content is an important component in oil palm plantation soil. The utilization of Near Infrared Sprctroscopy (NIRS) is an alternative that could replace laboratory test in order to analyze and provide information about measurement of potassium fertilizer content in oil palm soil. . This study aims to examine and evaluate NIRS technology as a faster and proper method in predicting rice moisture content by  Support Vector Regression (SVR) method and determining the best and accurate spectrum correction method to predict rice water content using Standard Normal Variate (SNV) pretreatment Multiplicative Spectral Correction (MSC) and  combination of both. This study used 100 soil samples with a wavelength of 350nm - 2500nm. Data processing using R software®  version 4.4.1. The results showed the prediction of the radial basis kernel SVR method, produced the best correction method in this study, namely Multiplicative Spectral Correction with an R2 value of 0.6025 and an RMSE of 0.0201.