Pengembangan Model Peramalan Jumlah Gangguan Jaringan Serat Optik Berbasis Deret Waktu

Prinandika, Arya Gading (2026) Pengembangan Model Peramalan Jumlah Gangguan Jaringan Serat Optik Berbasis Deret Waktu. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221280_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Penelitian ini membangun sistem pemeliharaan prediktif untuk meramalkan jumlah gangguan jaringan serat optik menggunakan dua pendekatan komparatif, yaitu pendekatan regresi numerik dan pendekatan klasifikasi ordinal. Pipeline data diawali dengan melakukan standardisasi log perbaikan melalui normalisasi nama Wilayah Telekomunikasi (WITEL) berbasis pencocokan fuzzy, pembersihan data, serta pengujian tiga skema agregasi waktu berupa bulanan, dua bulanan, dan tiga bulanan pada dua belas WITEL terpilih. Evaluasi performa model dilakukan secara ketat menggunakan metode walk-forward validation berbasis skema expanding window. Hasil eksperimen regresi menunjukkan bahwa model statistik AutoARIMA dan Non-Homogeneous Poisson Process (NHPP) secara kolektif mengungguli algoritma machine learning pada tujuh dari dua belas wilayah dalam menangkap pola data yang pendek. Pada pendekatan klasifikasi ordinal menggunakan algoritma K-Means untuk pelabelan kelas, model mampu menghasilkan performa kategori fair dengan nilai Cohen Kappa positif yang stabil pada empat wilayah operasional, yaitu Madiun, Pekanbaru, Pasuruan, dan Madura. Karakteristik data yang bervolume rendah dan bersifat zero-inflated menjadi faktor pembatas utama bagi performa model regresi kuantitatif, sehingga integrasi kedua pendekatan ini direkomendasikan secara komplementer untuk mendukung efisiensi alokasi sumber daya teknisi di lapangan.
================================================================================================================================
This study develops a predictive maintenance system to forecast optical fiber network disruptions using two comparative approaches, which are numerical regression and ordinal classification. The data pipeline begins by standardizing maintenance logs through fuzzy matching-based Wilayah Telekomunikasi (WITEL) name normalization, data cleaning, and testing three temporal aggregation schemes including monthly, bi-monthly, and tri-monthly intervals across twelve selected WITELs. Model performance evaluation is rigorously conducted using walk-forward validation with an expanding window scheme. The regression results demonstrate that statistical models, specifically Auto-ARIMA and the NonHomogeneous Poisson Process (NHPP), collectively outperform machine learning algorithms in seven out of twelve regions due to their capability in handling short data sequences. In the ordinal classification approach utilizing the K-Means algorithm for class labeling, the models achieve a fair performance category with stable positive Cohen Kappa values in four operational areas, namely Madiun, Pekanbaru, Pasuruan, and Madura. Low volume and zeroinflated data characteristics serve as the primary limiting factors for quantitative regression model accuracy, which makes the complementary integration of both approaches highly recommended to support efficient technician resource allocation in the field.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Peramalan Gangguan, Serat Optik, Pemeliharaan Prediktif, Walk-Forward Validation, Auto-ARIMA, NHPP, Klasifikasi Ordinal, Zero-Inflated Data, Disruption Forecasting, Optical Fiber, Predictive Maintenance, Ordinal Classification
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Arya Gading Prinandika
Date Deposited:	24 Jul 2026 01:28
Last Modified:	24 Jul 2026 01:28
URI:	http://repository.its.ac.id/id/eprint/136726

Actions (login required)

View Item