Dual-Temporal LSTM with Hybrid Attention for Airline Demand Forecasting
A recent study introduces a dual-stream Long Short-Term Memory (LSTM) model enhanced with attention mechanisms to better forecast airline passenger load factors. Existing models analyze booking data as a singular temporal dimension, focusing solely on either intra-flight accumulation or inter-flight trends, which overlooks valuable complementary insights. This innovative dual-stream method evaluates both a horizontal sequence (bookings leading up to departure) and a vertical sequence (flight patterns at specific days-before-departure intervals). By predicting load factors instead of total passenger numbers, the model mitigates issues arising from changes in aircraft types. It employs hybrid attention to prioritize significant features. This research, published on arXiv (2605.11569), seeks to overcome challenges in revenue management systems.
Key facts
- Dual-stream LSTM with attention framework proposed
- Processes intra-flight and inter-flight booking sequences
- Forecasts load factor instead of absolute passenger counts
- Addresses fragility from aircraft type changes
- Published on arXiv with ID 2605.11569
- Announcement type: new
- Aims to improve short-term demand forecasting for airline revenue management
- Hybrid attention mechanism integrates both temporal streams
Entities
Institutions
- arXiv