Deteksi Adulterasi Madu Berbasis Near-Infrared Spectroscopy (NIRS) dengan Model Machine Learning

‘Affifah, Putri Chentya Juni and Dr.Agr.Sc. Dimas Firmanda Al Riza,, ST., M.Sc and Prof. La Choviya Hawa,, STP., MP., Ph.D (2024) Deteksi Adulterasi Madu Berbasis Near-Infrared Spectroscopy (NIRS) dengan Model Machine Learning. Sarjana thesis, Universitas Brawijaya.

Abstract

Madu merupakan cairan alami yang memiliki cita rasa manis yang diproduksi oleh lebah madu dari nektar bunga atau bagian tanaman lainnya. Pelacakan keaslian madu dulu dilakukan menggunakan indera manusia, namun indera manusia memiliki keterbatasan karena madu asli dan madu palsu sulit dibedakan bila dilihat sekilas maupun dicium berdasarkan aromanya karena kedua madu mirip dan memiliki aroma beragam berdasarkan sumber madunya. Salah satu metode untuk mendeteksi madu palsu dan madu asli dengan akurasi tinggi, tidak merusak bahan, cepat, persiapan sampel yang sederhana, dan tidak membutuhkan bahan kimia, salah satunya adalah spektroskopi Near-Infrared (NIRS) yang diintegrasikan dengan model K-Nearest Neighbor (KNN) dan Support Vector Machine (SVM). Penelitian ini bertujuan untuk mendeteksi adulterasi madu berdasarkan pemodelan klasifikasi machine learning yang menghasilkan akurasi terbaik. Adulterasi madu dilakukan dengan menggunakan 10 merek sampel madu yang berasal dari pabrik madu, toko retail, dan toko online, serta bahan tambahan berupa sirup gula jagung tinggi fruktosa. Adulterasi madu dilakukan dengan konsentrasi 0%, 20%, 40%, 60%, dan 100%. Penelitian ini dilakukan melalui 2 tahap yaitu uji non-destruktif untuk mengetahui spektrum sampel madu menggunakan The SpectrapodTM pada rentang panjang gelombang 850-1.700 nm yang digabungkan dengan model KNN dan SVM, serta uji destruktif untuk mengetahui total padatan terlarut menggunakan Hand Held Refractometer dan kadar air sampel madu menggunakan Moisture Analyzer. Pada pemodelan klasifikasi dengan model KNN digunakan nilai k 3, 5, dan 7, dimana nilai k=3 menghasilkan nilai accuracy terbaik dengan nilai training accuracy sebesar 0,90 dan nilai testing accuracy sebesar 0,83. Sementara pemodelan klasifikasi dengan SVM digunakan kernel linear, polinomial, dan RBF (Radial Basis Function), dimana kernel polinomial menghasilkan nilai training accuracy tertinggi sebesar 0,87 dan nilai testing accuracy sebesar 0,70.

English Abstract

Honey is a natural liquid with a sweet flavour that is produced by honeybees from the nectar of flowers or other plant parts. Tracking the authenticity of honey used to be done by using human senses, but human senses have limitations because real honey and fake honey are difficult to distinguish when seen at a glance or smelled based on odour because both honey look similar and have various odours based on the honey source. One method that can be used to detect fake and real honey with the advantages of high accuracy, non-destructive, fast, simple sample preparation, and no chemicals needed is Near-Infrared (NIRS) spectroscopy integrated with K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) models. This study aims to detect honey adulteration based on machine learning classification modelling that has the best accuracy. Honey adulteration was conducted using 10 brands of honey samples from honey factories, retail stores, and online stores. The additive used was high-fructose corn sugar syrup. Honey adulteration was conducted with concentrations of 0%, 20%, 40%, 60%, and 100%. This research was conducted through 2 stages, namely non-destructive tests to determine the spectrum of honey samples using The SpectrapodTM in the wavelength range 850-1.700 nm and destructive tests to determine total dissolved solids using Hand Held Refractometer and water content of honey samples using a Moisture Analyzer. In modelling the classification with the KNN model, k values of 3, 5, and 7 are used, where the value of k = 3 has the best accuracy value with a training accuracy value of 0,90 and a testing accuracy value of 0,83. While in modelling the classification with the SVM model using lineal, polynomial, and RBF (Radial Basis Function) kernels, where the polynomial kernel has the highest accuracy value with a training accuracy value of 0,87 and a test accuracy value of 0,70.

Item Type:	Thesis (Sarjana)
Identification Number:	052410
Uncontrolled Keywords:	Adulterasi madu, Spektroskopi Near-Infrared, KNN, SVM, Total padatan terlarut, Kadar air
Divisions:	Fakultas Teknologi Pertanian > Keteknikan Pertanian
Depositing User:	Unnamed user with username nova
Date Deposited:	06 Mar 2024 06:27
Last Modified:	06 Mar 2024 06:27
URI:	http://repository.ub.ac.id/id/eprint/216917

Text (DALAM MASA EMBARGO) Putri Chentya Juni `Affifah.pdf Restricted to Registered users only Download (4MB)

Actions (login required)

View Item