LAPORAN PENELITIAN MODEL TRANSFORMER UNTUK PREDIKSI JUMLAH DISTRIBUSI KOMODITAS ANGKUTAN BARANG

SISWANTO, JOKO and HAKIM, MUHAMMAD IMAN NUR and HUMAMI, FARIS and ARDIANTO, ARIF and PRASETYO, ILHAM BAGUS and NUGRAHA, VIKY DWI and HANDIANSYAH, NOOR ALIF and FADHILAH, MUHAMMAD and ILHAM, HAFIZAN ROFIQY and ARIFFANSYAH, YUNINDRA EKA (2026) LAPORAN PENELITIAN MODEL TRANSFORMER UNTUK PREDIKSI JUMLAH DISTRIBUSI KOMODITAS ANGKUTAN BARANG. Technical Report. POLITEKNIK KESELAMATAN TRANSPORTASI JALAN, TEGAL. (Submitted)

[img] Text (LAPORAN PENELITIAN)
LAPORAN PENELITIAN MODEL TRANSFORMER UNTUK PREDIKSI JUMLAH DISTRIBUSI KOMODITAS ANGKUTAN BARANG.pdf
Restricted to Registered users only
Download (1MB) | Request a copy

Abstract

Fluctuations in demand and the complexity of freight data on roads, which play a strategic role in goods distribution, demand more accurate predictions. A transformer-based freight volume prediction model with auto-tuning hyperparameters is proposed to improve accuracy using Bayesian Optimization (BO), Particle Swarm Optimization (PSO), and Population-Based Training (PBT). A three-year daily dataset is used, divided into training (80%) and testing (20%) data with MinMax Scaler normalization. Evaluation using MASE, R², and MSLE shows that PSO provides superior performance with high consistency, fast convergence, and residuals that meet the assumptions of normality and homoscedasticity. The integration of FEDformerPSO into a time-series freight volume prediction model produces more accurate 30-day short-term predictions with an initial fluctuation pattern followed by a stable upward trend. These results provide important insights for stakeholders in freight distribution and fleet allocation.

Item Type: Monograph (Technical Report)
Uncontrolled Keywords: freight transport, time-series prediction, transformer, PSO
Subjects: T Technology > TA Engineering (General). Civil engineering (General)
Divisions: Pusat Penelitian dan Pengabdian kepada Masyarakat > Pusat Penelitian dan Pengabdian kepada Masyarakat
Depositing User: PUSAT PENELITIAN DAN PENGABDIAN KEPADA MASYARAKAT PKTJ
Date Deposited: 04 Feb 2026 02:57
Last Modified: 04 Feb 2026 02:57
URI: http://eprints.pktj.ac.id/id/eprint/4235

Actions (login required)

View Item View Item
PKTJ Repository is powered by EPrints 3 which is developed by the School of Electronics and Computer Science at the University of Southampton. More information and software credits.