Pendugaan Kadar Patchouli Alkohol Pada Minyak Nilam Variasi Menggunakan Teknologi Near Infrared Reflectance Spectroscopy Dengan Metode Partial Least Square Regression

Abstrak. Minyak nilam merupakah salah satu minyak atsiri yang menyumbangkan devisa bagi negara. Komponen utama dalam minyak adalah Patchouli Alkohol (PA). Salah satu parameter yang menentukan kualitas minyak nilam adalah kadar PA. Semakin tinggi kadar PA maka  kualitasnya semakin baik dan harganya semakin tinggi. Tujuan dari penelitian ini adalah memprediksi kadar PA minyak nilam Aceh hasil fraksinasi dengan cepat dan tepat menggunakan teknologi NIRS dengan metode partial least square regression (PLSR) menggunakan pretreatment Mean Normalization (MN). Hasil penelitian menunjukkan bahwa PLSR mampu memprediksi kadar patchouli alkohol dengan menghasilkan model yang tergolong good prediction accuracy. Prediksi model akurasi terbaik dari perlakuan pretreatment MN dengan hasil dengan latent variable 5, nilai residual predictive deviation (RPD) sebesar 2,71, nilai koefisien determinasi (R2) sebesar 0,85, nilai koefisien korelasi (r) sebesar 0,92, nilai root mean square error calibration (RMSEC) sebesar 4,28.Estimation of Patchouli Alcohol Content in Variation of Patchouli Oil Using Near Infrared Reflectance Spectroscopy Technology with Partial Least Square Regression MethodAbstract. Patchouli oil is one type of essential oils that protects foreign exchange for the country. The main component in the oil is Patchouli Alcohol (PA). One of the parameters that determine the quality of patchouli oil is the PA content. The higher the PA content, the better the quality and the higher the price. The purpose of this study was to predict the PA content of fractionated Aceh patchouli oil quickly and precisely using NIRS technology with the partial least square regression (PLSR) method using Mean Normalization (MN) pretreatment. The results showed that PLSR was able to predict patchouli alcohol levels by producing a model that was classified as having good predictive accuracy. The best accuracy prediction model of the MN pretreatment handling with results with a latent variable of 5, the residual predictive deviation (RPD) value about 2,71, the coefficient of determination (R2) is 0,85, correlation coefficient (r) 0,92, and the value root mean square error calibration (RMSEC) of 4,28.