TripCast: Pre-training of Masked 2D Transformers for Trip Time Series Forecasting
This work addresses forecasting problems for the tourism industry, offering a novel approach to handle 2D time series data.
The authors tackled the challenge of forecasting trip time series with a 2D structure in the tourism industry by proposing TripCast, a pre-trained model using masked 2D transformers, which notably outperformed state-of-the-art baselines in in-domain scenarios and showed strong scalability and transferability in out-domain scenarios.
Deep learning and pre-trained models have shown great success in time series forecasting. However, in the tourism industry, time series data often exhibit a leading time property, presenting a 2D structure. This introduces unique challenges for forecasting in this sector. In this study, we propose a novel modelling paradigm, TripCast, which treats trip time series as 2D data and learns representations through masking and reconstruction processes. Pre-trained on large-scale real-world data, TripCast notably outperforms other state-of-the-art baselines in in-domain forecasting scenarios and demonstrates strong scalability and transferability in out-domain forecasting scenarios.