TDCM: Transport Destination Calibrating Based on Multi-Task Learning
Abstract
Accurate location and address of destination are critical for bulk commodity transportation, which determines the service quality of the logistics applications such as transport task dispatching and route planning. But due to manual input errors of the operators and dynamic changes of the destination’s location, the address of destination is not always correct and complete. To tackle this issue, we propose T ransport D estination C alibration framework based on M ulti-task learning, called TDCM . To correctly pinpoint the locations of destinations that are close to each other but differ in size, we cluster stay points to get stay areas and then merge them based on road turn-off location to obtain stay hotspots. Further, to precisely recognize the transport destination for each waybill, we devise an end-to-end multi-task destination matching model by incorporating with an attention mechanism. It can identify all destinations’ instances and meanwhile can match them with the corresponding waybills’ addresses respectively. Experimental results on real-world steel logistics data demonstrate the effectiveness and superiority of TDCM .
Cite
Text
Wu et al. "TDCM: Transport Destination Calibrating Based on Multi-Task Learning." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2023. doi:10.1007/978-3-031-43430-3_17Markdown
[Wu et al. "TDCM: Transport Destination Calibrating Based on Multi-Task Learning." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2023.](https://mlanthology.org/ecmlpkdd/2023/wu2023ecmlpkdd-tdcm/) doi:10.1007/978-3-031-43430-3_17BibTeX
@inproceedings{wu2023ecmlpkdd-tdcm,
title = {{TDCM: Transport Destination Calibrating Based on Multi-Task Learning}},
author = {Wu, Tao and Zhu, Kaixuan and Mao, Jiali and Yang, Miaomiao and Zhou, Aoying},
booktitle = {European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases},
year = {2023},
pages = {276-292},
doi = {10.1007/978-3-031-43430-3_17},
url = {https://mlanthology.org/ecmlpkdd/2023/wu2023ecmlpkdd-tdcm/}
}