DeepExpress: Heterogeneous and Coupled Sequence Modeling for Express Delivery Prediction

The prediction of express delivery sequence, i.e., modeling and estimating the volumes daily incoming outgoing parcels for delivery, is critical online business, logistics, positive customer experience, specifically resource allocation optimization promotional activity arrangement. A precise estimate consumer requests has to involve sequential factors such as shopping behaviors, weather conditions, events, business campaigns, their couplings. Despite that various methods have integrated external features enhance effects, extant works fail address complex feature-sequence couplings in following aspects: weaken inter-dependencies when processing heterogeneous data ignore cumulative evolving situation coupling relationships. To these issues, we propose DeepExpress—a deep-learning-based sequence model, which extends classic seq2seq framework learn DeepExpress leverages an learning, a carefully designed feature representation, novel joint training attention mechanism adaptively handle heterogeneity issues capture accurate prediction. Experimental results on real-world demonstrate proposed method outperforms both shallow deep baseline models.