Analisis Karakteristik Spektrum NIRS dan Identifikasi Asal Daerah Biji Kopi Robusta Menggunakan Metode PLS-DA
Abstract
Varietas kopi yang banyak dibudidayakan di Indonesia salah satunya adalah kopi robusta. Beragamnya indikasi geografis di Indonesia membuat varietas kopi robusta juga beragam, sehingga memperbesar peluang pemalsuan produk dalam pendistribusian. Penelitian ini bertujuan untuk menganalisis karakteristik spektrum biji kopi robusta dari Sumatera Barat, Jawa Barat, dan Jambi secara non destruktif menggunakan Near Infrared Spectroscopy (NIRS) serta membangun model Partial Least Square Discriminant Analysis (PLS-DA) untuk mengidentifikasi keaslian dari biji kopi robusta berdasarkan asal daerahnya. Penelitian ini menggunakan biji kopi robusta yang masih dalam bentuk green bean yang diambil dari 3 daerah, yaitu Sumatera Barat, Jawa Barat, dan Jambi. Total sampel yang digunakan berjumlah 180 sampel, sebanyak 120 sampel dipakai untuk kalibrasi dan 60 sampel untuk evaluasi. Spektra reflektansi biji kopi diukur menggunakan NIRS pada panjang gelombang 1000-2500 nm. Setelah pengukuran reflektansi, sampel biji kopi dilakukan analisis komposisi kimia menggunakan metode proksimat. Spektra reflektansi diolah dengan berbagai pretreatment (Savitzky-Golay First Derivative (SG1D), Standard Normal Variate (SNV), Multiplicative Scatter Correction (MSC), dan pretreatment kombinasi). Model PLS-DA yang dibangun menggunakan pretreatment Savitzky-Golay First Derivative baik tunggal maupun kombinasi (SG1D+SNV, SG1D+MSC, SG1D+SNV+MSC) mampu memprediksi biji kopi sesuai asal daerahnya dengan sangat baik, dengan nilai akurasi, sensitivitas, spesifisitas, dan presisi, masing-masing mencapai 99%; 98%; 99% dan 98%. Hasil ini menunjukkan bahwa biji kopi robusta yang berbeda asal daerahnya dapat diklasifikasikan atau diidentifikasi keasliannya secara non destruktif menggunakan NIRS. One of the coffee varieties that is widely cultivated in Indonesia is robusta coffee. The variety of geographical indications in Indonesia means that robusta coffee varieties are also diverse, thereby increasing the opportunity for product counterfeiting in distribution. This research aims to analyze the spectral characteristics of robusta coffee beans from West Sumatra, West Java, and Jambi non-destructively using Near Infrared Spectroscopy (NIRS) and build a Partial Least Square Discriminant Analysis (PLS-DA) model to identify the authenticity of robusta coffee beans based on area of origin. This research uses robusta coffee beans which are still in green bean form taken from 3 regions, namely West Sumatra, West Java, and Jambi. The total samples used were 180 samples, 120 samples were used for calibration, and 60 samples for evaluation. The reflectance spectrum of coffee beans was measured using NIRS at a wavelength of 1000-2500 nm. After measuring the reflectance, the coffee bean samples were subjected to chemical composition analysis using the proximate method. Reflectance spectra are processed with various pretreatments (Savitzky-Golay First Derivative (SG1D), Standard Normal Variate (SNV), Multiplicative Scatter Correction (MSC), and pretreatments combination). The PLS-DA model built using Savitzky-Golay First Derivative pretreatment, both single and combined (SG1D+SNV, SG1D+MSC, SG1D+SNV+MSC), is capable of predicting coffee beans according to their origin very well, with accuracy, sensitivity, specificity, and precision values reaching 99%, 98%, 99%, and 98%, respectively. These results show that robusta coffee beans from different regions of origin can be classified or identified for their authenticity non-destructively using NIRS.