DefinePK

Abstract

Near infrared (NIR) spectroscopic analysis has been widely evaluated in various areas due to its potential to be an alternative of numerous conventional measurement approaches that are time consuming, expensive, or destructive. This study evaluated the feasibility of adaptive linear neuron (Adaline) to be implemented as a variable selection approach to identify effective NIR wavelengths that can be used to predict the soil organic matter (SOM) so that a parsimonious model can be built. Adaline was optimized using its optimal learning rate and training adaptation cycles. After that, the effective wavelengths were identified based on the weight values of the best Adaline. The best predictive accuracy was achieved by the proposed Adaline that used 40 of the total 891 wavelengths with the root mean square error of prediction (RMSEP) and correlation coefficient of prediction (rp) of 2.163% and 0.9849, respectively. Findings show that the proposed variable selection approach by means of Adaline is capable of producing a parsimonious model that was able to predict the soil organic matter with better accuracy.