融合语义向量与微链结构特征的购物出行模式识别

Identifying Shopping Travel Patterns by Integrating Semantic Vectors and Micro-Chain Structural Features

摘要:

聚焦城市居民出行行为中的前后活动关联、出行方式及关键时间节点等特征，围绕购物活动构建微链结构，提出一种融合结构化出行特征与语义向量的购物出行模式识别框架，引入词嵌入与整体语义表达表征整条出行链的语义信息，结合时间编码与活动顺序特征，通过K-means++聚类方法挖掘城市居民购物出行链中隐含的行为模式。以上海居民购物出行为实证对象，聚类分析识别出5类具有显著差异的购物出行模式。对比结果显示，该方法在聚类清晰度与结果可解释性方面均优于传统基于出行属性的编码方式。研究不仅为理解城市居民购物出行行为提供了细粒度视角，也为城市出行行为研究提供了一种具有推广价值的特征构建与聚类识别框架。

Abstract:

To better capture the characteristics of urban residents’ travel behavior， such as the connections between pre- and post-activities， travel modes， and key time nodes， this paper constructs a “micro-chain” structure centered on shopping activities and proposes a shopping travel pattern recognition framework that integrates structured travel features with semantic vectors. The framework leverages Word2Vec and SIF to encode the semantic information of entire travel chains and combines this with time encoding and activity sequence features. Using the K-means++ clustering method， it uncovers latent behavioral patterns in shopping-related travel chains of urban residents. Applying the framework to shopping travel data from Shanghai residents， five distinct shopping travel patterns with significant differences are identified. Comparative analyses show that the proposed method outperforms traditional travel attribute-based encoding approaches in both clustering clarity and interpretability. This paper provides a fine-grained perspective on urban shopping travel behavior and offers a transferable feature construction and clustering framework for urban travel behavior research.