Identifikasi Kualitas Minuman Legen Berbasis Multi-Sensor Data Fusion dan Machine Learning

Khakim, Lukmanul (2026) Identifikasi Kualitas Minuman Legen Berbasis Multi-Sensor Data Fusion dan Machine Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6022232013-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Legen merupakan minuman tradisional dari air nira pohon siwalan (Borassus flabellifer L.) dan juga dimanfaatkan sebagai bahan baku olahan pangan bernilai ekonomi tinggi, namun legen mudah mengalami penurunan kualitas akibat fermentasi mikroba selama penyimpanan. Penentuan kualitas legen secara langsung tidak akurat dan tidak konsisten sedangkan analisis laboratorium yang memakan waktu serta biaya. Pada penelitian ini, penentuan label kualitas legen dilakukan secara kualitatif melalui uji organoleptik oleh panel ahli dengan perlakuan lama penyimpanan sampel, sehingga diperoleh tiga kategori kualitas, yaitu baik, sedang, dan buruk. Penelitian ini mengusulkan sistem identifikasi kualitas legen berbasis multi-sensor data fusion yang mengintegrasikan deret sensor gas, deret sensor potensiometri, dan sensor warna RGB dengan pendekatan machine learning. Deret sensor gas yang digunakan terdiri dari TGS2602, MQ-135, dan MQ-3, pemilihan ketiga sensor gas tersebut didasarkan pada karakteristik senyawa volatil pada legen. Deret potensiometri terdiri dari elektroda stainless steel, titanium dan tembaga yang dikombinasikan dengan elektroda Ag/AgCl. Pengujian dilakukan menggunakan dua pendekatan data fusion, yaitu low-level fusion dan mid-level fusion, serta empat algoritma klasifikasi: RF, XGBoost, SVM, dan NN. Hasil penelitian menunjukkan bahwa kombinasi tiga jenis sensor memberikan kinerja terbaik, dengan sensor gas sebagai sensor paling dominan. Secara umum, low-level fusion menghasilkan akurasi rata-rata lebih tinggi sebesar 87,7% dibandingkan mid-level fusion sebesar 86,3%. Model Neural Network mencapai akurasi tertinggi sebesar 94,6% pada low-level fusion, sedangkan XGBoost menunjukkan performa kompetitif dengan akurasi hingga 94,5% pada mid-level fusion. Hasil ini menunjukkan bahwa sistem ini efektif untuk mengidentifikasi kualitas minuman legen secara cepat, praktis, dan berbiaya relatif rendah.
=================================================================================================================================
Legen is a traditional drink made from the sap of the siwalan palm (Borassus flabellifer L.) and is also used as a raw material for high-value food products. However, legen easily deteriorates in quality due to microbial fermentation during storage. Direct determination of legen quality is inaccurate and inconsistent, while laboratory analysis is time-consuming and costly. In this study, the quality label of legen was determined qualitatively through organoleptic testing by a panel of experts with sample storage duration treatment, resulting in three quality categories, namely good, moderate, and poor. This study proposes a multi-sensor data fusion-based legen quality identification system that integrates a gas sensor array, a potentiometric sensor array, and an RGB color sensor with a machine learning approach. The gas sensor series used consisted of TGS2602, MQ-135, and MQ-3. The selection of these three gas sensors was based on the characteristics of volatile compounds in legen. The potentiometric series consisted of stainless steel, titanium, and copper electrodes combined with Ag/AgCl electrodes. Testing was conducted using two data fusion approaches, namely low-level fusion and mid-level fusion, as well as four classification algorithms: RF, XGBoost, SVM, and NN. The results showed that the combination of the three types of sensors provided the best performance, with the gas sensor being the most dominant. In general, low-level fusion produced a higher average accuracy of 87.7% compared to mid-level fusion at 86.3%. The Neural Network model achieved the highest accuracy of 94.6% in low-level fusion, while XGBoost showed competitive performance with an accuracy of up to 94.5% in mid-level fusion. These results indicate that this system is effective for identifying the quality of legen drinks quickly, practically, and at a relatively low cost.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Legen, Multi-Sensor, Machine learning, Pangan, Food, Legen, Multi-Sensor, Machine learning
Subjects:	T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models. T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK351 Electric measurements. T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7871.674 Detectors. Sensors
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User:	Lukmanul Khakim
Date Deposited:	21 Jan 2026 07:18
Last Modified:	21 Jan 2026 07:18
URI:	http://repository.its.ac.id/id/eprint/129941

Actions (login required)

View Item