Answers are grounded in this paper and cite their sources.
Home-based basic trip-chain prediction jointly characterizes trip purpose, destination, time, and travel mode, and is fundamental to travel behavior modeling and fine-grained demand analysis. Existing models often focus on single tasks or parallel multi-task outputs, but provide limited support for high-cardinality traffic analysis zone (TAZ) destination candidate generation, chained task propagation, and utility semantics in deep models. This study proposes the Basic Trip Neural Network (BTrip-NN) to predict trip purpose, destination TAZ, departure and arrival time periods, and primary and secondary travel modes. The model uses a Purpose → Destination → Time → Mode cascade, learns TAZ spatial embeddings with a Hybrid Attention Graph Convolutional Network (HA-GCN), and retains utility-probability semantics through a dynamic parameter utility layer. Using a household travel survey from the central urban area of Chongqing, the model is evaluated over 1,604 destination TAZs. BTrip-NN achieves a destination Top-10 coverage of 0.44. After incorporating observed fixed workplace and school-location priors at the prediction stage, anchor-aware Top-10 coverage reaches 0.88. The dynamic parameter utility layer yields a 10.37% Top-10 gain, with a bootstrap 95% confidence interval (CI) of [0.09, 0.11]. Compared with multinomial logit (MNL), tree-based models, graph convolutional network (GCN), graph attention network (GAT), sequential models, and standard multi-task networks, BTrip-NN shows more stable performance in destination candidate coverage and balanced recognition of non-spatial tasks. The results suggest that cascaded task propagation, spatial graph representation, and dynamic utility adjustment help improve destination candidate coverage and multi-task recognition in TAZ-level basic trip-chain prediction.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: