DETEKSI PEMALSUAN MINYAK ATSIRI BERBASIS E-NOSE DAN MACHINE LEARNING DENGAN EKSTRAKSI FITUR MULTI-WINDOW

Putri Aurelia, . (2026) DETEKSI PEMALSUAN MINYAK ATSIRI BERBASIS E-NOSE DAN MACHINE LEARNING DENGAN EKSTRAKSI FITUR MULTI-WINDOW. Sarjana thesis, UNIVERSITAS NEGERI JAKARTA.

	Text Cover.pdf Download (1MB)
	Text BAB 1.pdf Download (329kB)
	Text BAB 2.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text BAB 3.pdf Restricted to Registered users only Download (382kB) | Request a copy
	Text BAB 4.pdf Restricted to Registered users only Download (496kB) | Request a copy
	Text BAB 5.pdf Restricted to Registered users only Download (309kB) | Request a copy
	Text Daftar Pustaka.pdf Download (231kB)
	Text Lampiran_compressed.pdf Restricted to Registered users only Download (53kB) | Request a copy

Abstract

Pemalsuan minyak atsiri merupakan salah satu permasalahan yang dapat menurunkan kualitas dan nilai ekonomi produk sehingga diperlukan metode deteksi yang cepat, akurat, dan nondestruktif. Electronic Nose (E-Nose) merupakan teknologi yang mampu mendeteksi senyawa volatil berdasarkan pola respons sensor gas. Namun, penggunaan keseluruhan sinyal dalam proses ekstraksi fitur belum tentu mampu merepresentasikan dinamika respons sensor secara optimal. Penelitian ini bertujuan untuk menganalisis karakteristik sinyal E-Nose pada minyak atsiri temulawak asli dan minyak atsiri temulawak yang dipalsukan, menerapkan ekstraksi fitur multi-window untuk memperoleh representasi sinyal yang lebih informatif, serta mengevaluasi kinerja beberapa algoritma machine learning dalam mendeteksi pemalsuan minyak atsiri temulawak. Data penelitian diperoleh menggunakan E-Nose yang terdiri atas 10 sensor Metal Oxide Semiconductor (MOS). Sinyal sensor diproses melalui baseline correction dan cropping, kemudian dilakukan ekstraksi fitur menggunakan pendekatan multi-window. Fitur yang diperoleh selanjutnya diseleksi menggunakan Mutual Information sebelum dilakukan klasifikasi menggunakan Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), dan Extreme Gradient Boosting (XGBoost). Hasil penelitian menunjukkan bahwa pendekatan multi-window mampu memberikan representasi karakteristik sinyal yang lebih informatif dan meningkatkan kinerja klasifikasi pada beberapa algoritma dibandingkan pendekatan single-window. Model SVM dengan fitur hasil seleksi multi-window memberikan kinerja terbaik dengan akurasi sebesar 95%, dengan nilai macro-precision, macro-recall, dan macro-F1 score masing-masing sebesar 96%, 95%, dan 95%. Hasil tersebut menunjukkan bahwa kombinasi E-Nose, ekstraksi fitur multi-window, seleksi fitur Mutual Information, dan machine learning berpotensi menjadi pendekatan yang efektif untuk mendukung deteksi pemalsuan minyak atsiri temulawak secara cepat dan nondestruktif. Kata kunci: Electronic Nose, minyak atsiri temulawak, ekstraksi fitur multi-window, Mutual Information, machine learning, pemalsuan minyak atsiri. ****** Essential oil adulteration is a major issue that can reduce product quality and economic value, creating the need for rapid, accurate, and non-destructive detection methods. Electronic Nose (E-Nose) is a technology capable of detecting volatile compounds based on gas sensor response patterns. However, the use of the entire sensor signal for feature extraction may not adequately represent the temporal dynamics of sensor responses. This study aims to analyze the response characteristics of an E-Nose for authentic and adulterated Curcuma xanthorrhiza essential oil, implement a multi-window feature extraction approach to obtain a more informative signal representation, and evaluate the performance of several machine learning algorithms for essential oil adulteration detection. The study used sensor response data acquired from an E-Nose equipped with ten Metal Oxide Semiconductor (MOS) sensors. The sensor signals were preprocessed through baseline correction and signal cropping, followed by feature extraction using a multi-window approach. The extracted features were then selected using Mutual Information before classification using Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and Extreme Gradient Boosting (XGBoost). The results show that the multi-window approach provides a more informative representation of sensor signal characteristics and improves classification performance for several algorithms compared with the single-window approach. The SVM model using the selected multi-window features achieved the best performance, with an accuracy of 95%, macro-precision of 96%, macro-recall of 95%, and macro-F1 score of 95%. These results indicate that the combination of E-Nose, multi-window feature extraction, Mutual Information feature selection, and machine learning has the potential to be an effective approach for supporting rapid and non-destructive detection of Curcuma xanthorrhiza essential oil adulteration. Keywords: Electronic Nose, Curcuma xanthorrhiza essential oil, multi-window feature extraction, Mutual Information, machine learning, essential oil adulteration.

Item Type:	Thesis (Sarjana)
Additional Information:	1). Dr. Bambang Heru Iswanto, M.Si. ; 2). Dr. Fajar Hardoyono, M.Sc.
Subjects:	Sains > Fisika
Divisions:	FMIPA > S1 Fisika
Depositing User:	Users 36181 not found.
Date Deposited:	12 Aug 2026 03:57
Last Modified:	12 Aug 2026 03:57
URI:	http://repository.unj.ac.id/id/eprint/70165

Actions (login required)

View Item