Optimasi Long Short-Term Memory Menggunakan Algoritma Genetika untuk Perencanaan Kapasitas Commuter Line

Ardiansyah, Muchammad Aqik (2026) Optimasi Long Short-Term Memory Menggunakan Algoritma Genetika untuk Perencanaan Kapasitas Commuter Line. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 603224108-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Fluktuasi harian permintaan penumpang yang dinamis dan tidak terprediksi merupakan tantangan utama dalam perencanaan kapasitas sistem kereta api perkotaan. Penelitian ini menguji apakah integrasi data Google Trends sebagai variabel eksogen dapat meningkatkan akurasi prakiraan penumpang Commuter Line Jabodetabek. Model Long Short-Term Memory (LSTM) diterapkan dengan optimasi hyperparameter melalui penyetelan manual maupun Genetic Algorithm (GA). Menggunakan data harian periode Januari 2021 hingga Juli 2025, empat skenario dievaluasi: Model 1 (LSTM manual tanpa Google Trends), Model 2 (LSTM manual dengan Google Trends), Model 3 (LSTM-GA tanpa Google Trends), dan Model 4 (LSTM-GA dengan Google Trends). Analisis korelasi silang mengonfirmasi hubungan yang kuat antara tren pencarian daring dan volume penumpang. Model 4 mencapai kinerja terbaik (RMSE 104.506; MAPE 8,83%), disusul oleh Model 3 (RMSE 110.927; MAPE 10,16%). Temuan menunjukkan bahwa optimasi berbasis GA berdampak signifikan pada perbaikan galat (error) prediksi, sementara penggunaan variabel Google Trends terbukti efektif dalam mendukung proses pelatihan model yang good-fit. Model 4 direkomendasikan sebagai solusi optimal karena berhasil mensinergikan GA untuk presisi parameter dan Google Trends sebagai indikator perilaku masyarakat. Sementara itu, Model 3 berfungsi sebagai mekanisme cadangan operasional (fail-safe) yang andal apabila terjadi kendala pada ketersediaan data eksternal. Studi ini menyediakan alat pendukung keputusan bagi operator untuk mengantisipasi lonjakan penumpang dan mengoptimalkan perencanaan kapasitas operasional secara lebih efektif.
=====================================================================================================================================
Daily fluctuations in passenger demand, which are dynamic and unpredictable, represent a primary challenge in urban railway capacity planning. This study examines whether integrating Google Trends data as an exogenous variable can enhance the accuracy of passenger forecasts for the Jabodetabek Commuter Line. A Long Short-Term Memory (LSTM) model is implemented with hyperparameter optimization through both manual tuning and Genetic Algorithms (GA). Using daily data from January 2021 to July 2025, four scenarios were evaluated: Model 1 (Manual LSTM without Google Trends), Model 2 (Manual LSTM with Google Trends), Model 3 (LSTM-GA without Google Trends), and Model 4 (LSTM-GA with Google Trends). Cross-correlation analysis confirmed a strong relationship between online search trends and passenger volume. Model 4 achieved the best performance (RMSE 104,506; MAPE 8.83%), followed by Model 3 (RMSE 110,927; MAPE 10.16%). Findings indicate that GA-based optimization significantly impacts error reduction, whereas the inclusion of Google Trends variables effectively supports a good-fit training process. Model 4 is recommended as the optimal solution as it successfully synergizes GA for parameter precision and Google Trends as an indicator of public behavior. Meanwhile, Model 3 serves as a reliable operational fail-safe mechanism in the event of external data availability issues. This study provides a decision-support tool for operators to anticipate passenger surges and optimize operational capacity planning more effectively.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Perencanaan Kapasitas, Jalur Komuter, Google Trends, Hybrid GA-LSTM, Peramalan Nonlinier Capacity Planning, Commuter Line, Google Trends, Hybrid GA-LSTM, Nonlinear Forecasting
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T57.84 Heuristic algorithms.
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Muchammad Aqik Ardiansyah
Date Deposited:	28 Jan 2026 06:17
Last Modified:	28 Jan 2026 06:17
URI:	http://repository.its.ac.id/id/eprint/130775

Actions (login required)

View Item