<![CDATA[AbstractGreen coffee has a variety of benefits that are good for the human body, because of the antioxidants content like chlorogenic acid and trigonelline as well as caffeine as a central nervous stimulant. Commonly, determination of these chemical contents was done by the chemical method that less efficient in terms of time, cost and sample preparation. NIR Spectroscopy has been applied as an alternative method for prediction of these chemical contents but the accuracy is not quite accurate. In this research, Kubelka-Munk Model was applied to increase the accuracy of NIRS for prediction of these chemical content of bondowoso green coffee powder. The sample of coffee was grounded into particle size of 355 and 150 μm, and the reflectance of the sample (30 gram) were measured by FT-NIRS in the wavelength of 1000-2500 nm. Furthermore, the chemical content of the samples were determined by Liquid Chromatography Mass Spectrometry (LCMS). The obtained spectrum was transformed to absorbance (Log 1/R) and K/S of Kubelka-Munk model. Data pretreatment such as standard normal variate (SNV), second derivative (dg2) and their combination was also done to increase accuracy of NIRS prediction. The calibration and validation of processed NIR spectra and chemical content were carried out using Partial Least Square (PLS). The results show that K/S of Kubelka-Munk model was continued with data pretreatment dg2 on the 150 μm particle size of coffee powder giving the best prediction of caffeine, trigonelline and CGA of Bondowoso green coffee powder by NIRS (R2 > 0.98; RPD >5.31; CV < 1.07%).AbstrakGreen coffee memberikan berbagai manfaat bagi kesehatan tubuh manusia, karena kandungan antioksidan seperti asam klorogenat dan trigonelin serta kafein sebagai perangsang sistem syaraf pusat. Umumnya, penentuan kandungan ini dilakukan dengan metode kimia yang kurang efisien dalam waktu, mahal dan perlu persiapan sampel. NIR Spectroscopy telah diterapkan sebagai metode alternatif untuk prediksi kandungan ini, namun hasilnya tidak terlalu akurat. Dalam penelitian ini, model Kubelka-Munk diterapkan untuk peningkatan akurasi NIRS dalam memprediksi kandungan kimia bubuk green coffee Bondowoso. Sampel kopi digiling pada ukuran partikel 355 dan 150 μm, dan reflektan sampel (30 gram) diukur dengan FT-NIRS pada panjang gelombang 1000-2500 nm. Selanjutnya, pengukuran kandungan kimia sampel dilakukan dengan Liquid Chromatography Mass Spectrometry (LCMS). Spektrum yang diperoleh ditransformasi ke absorban (Log 1/R) dan K/S dari model Kubelka-Munk. Data pretreatment seperti standard normal variate (SNV), second derivative (dg2) dan kombinasinya juga dilakukan untuk meningkatkan akurasi prediksi NIRS. Kalibrasi dan validasi spektra NIR terolah dengan data kimia dilakukan menggunakan Partial Least Square (PLS). Hasil penelitian menunjukkan bahwa K/S dari model Kubelka-Munk dilanjutkan dengan data pretreatment dg2 pada ukuran partikel bubuk kopi 150 μm memberikan prediksi terbaik untuk penentuan kandungan kafein, trigonelin dan CGA dari bubuk green coffee Bondowoso dengan NIRS (R2 > 0.98; RPD >5.31; CV < 1.07%).]]>
