Implementasi Model Klasifikasi dengan Metode Shapelet pada Data Time Series Multivariat Internet of Things

Farrel, Amanda Illona (2026) Implementasi Model Klasifikasi dengan Metode Shapelet pada Data Time Series Multivariat Internet of Things. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221056_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Deteksi kerusakan pada data deret waktu multivariat bertujuan mengidentifikasi kondisi yang menyimpang dari pola normal suatu sistem. Proses ini penting dalam pemantauan kondisi mesin industri karena mampu memberikan peringatan dini terhadap potensi kerusakan. Pada Internet of Things (IoT) industri, data sensor bersifat multivariat dan berdimensi tinggi, sehingga menimbulkan tantangan berupa kompleksitas dan redundansi fitur, serta ketidakseimbangan kelas. Penelitian ini menerapkan Shapelet Transform untuk ekstraksi fitur yang dikombinasikan dengan augmentasi data dan algoritma klasifikasi konvensional, yaitu Random Forest (RF), Support Vector Machine (SVM), dan K-Nearest Neighbors (KNN). Dataset yang digunakan adalah Industrial IoT Fault Detection Dataset dengan tiga sensor, yaitu Vibration, Temperature, dan Pressure, serta tiga kelas yang tidak seimbang. Pengujian dilakukan melalui tiga skenario, yaitu komparasi fitur, optimasi panjang shapelet, dan hyperparameter tuning. Hasil menunjukkan kombinasi Shapelet Transform dengan Random Forest pada panjang shapelet 4–7 memperoleh performa terbaik, dengan Accuracy 0,9555, Precision macro 0,9602, Recall macro 0,9561, F1-score macro 0,9577, dan ROC-AUC macro 0,9932 setelah hyperparameter tuning. Augmentasi data meningkatkan F1-score macro dari 0,8601 menjadi 0,9468 sekaligus meningkatkan kemampuan deteksi kelas minoritas Overheating. Penelitian ini menunjukkan bahwa Shapelet Transform efektif menghasilkan representasi fitur yang ringkas dan diskriminatif untuk deteksi kerusakan pada data deret waktu multivariat IoT.
======================================================================================================================================
Fault detection in multivariate time series data aims to identify conditions that deviate from a system's normal patterns. This process is important in industrial machine condition monitoring as it enables early warning of potential failures. In industrial IoT environments systems, sensor data are multivariate and high-dimensional, posing challenges such as feature complexity, feature redundancy, and class imbalance. This study applies the Shapelet Transform for feature extraction, combined with data augmentation and conventional classification algorithms, namely Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN). The dataset used is the Industrial IoT Fault Detection Dataset, comprising three sensors—Vibration, Temperature, and Pressure—and three imbalanced classes. The evaluation was conducted through three scenarios: feature comparison, shapelet length optimization, and hyperparameter tuning. The results show that the combination of Shapelet Transform and Random Forest with a shapelet length of 4–7 achieved the best performance, with an Accuracy of 0.9555, macro Precision of 0.9602, macro Recall of 0.9561, macro F1-score of 0.9577, and macro ROC-AUC of 0.9932 after hyperparameter tuning. Data augmentation improved the macro F1-score from 0.8601 to 0.9468 and enhanced the detection of the minority Overheating class. This study shows that the Shapelet Transform effectively produces a compact and discriminative feature representation for fault detection in multivariate IoT time series data.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Internet of Things, Klasifikasi Deret Waktu, Multivariat, Multivariate, Predictive Maintenance, Shapelet Transform, Time Series Classification
Subjects:	Q Science > Q Science (General) > Q337.5 Pattern recognition systems Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA336 Artificial Intelligence
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Amanda Illona Farrel
Date Deposited:	22 Jul 2026 02:36
Last Modified:	22 Jul 2026 02:36
URI:	http://repository.its.ac.id/id/eprint/136420

Actions (login required)

View Item