Limit Management in Network Traffic Forecasting via Asymmetric Loss Function A?g Trafi?gi Tahminlemesinde Asimetrik Kayip Fonksiyonu ile Limit Yönetimi
34th Signal Processing and Communications Applications Conference, SIU 2026, İstanbul, Türkiye, 7 - 10 Temmuz 2026, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/siu71813.2026.11636729
- Basıldığı Şehir: İstanbul
- Basıldığı Ülke: Türkiye
- Anahtar Kelimeler: asymmetric loss function, network traffic, NeuralProphet, PELT, rolling Z-score, time series forecasting
- Orta Doğu Teknik Üniversitesi Adresli: Evet
Özet
This study focuses on the unpredictability of capacity expansion requirements driven by usage growth in telecommunication networks. In contrast to reactive methods, an end-to-end forecasting approach is proposed to enable intervention before bottlenecks occur. Anomalies in the dataset were mitigated using a rolling Z-score, and the current network characteristics were identified via the PELT algorithm. The NeuralProphet architecture was employed for the forecasting process, integrating an asymmetric loss function that heavily penalizes undershoots to ensure proactive capacity management. Evaluations conducted on real-world network traffic scenarios demonstrate that the proposed method increases the success rate of forecasting future capacity needs within a safety margin while maintaining the overall low error margin.