Community Search in Time-dependent Road-social Attributed Networks

TL;DR

This paper addresses semantic-spatial aware k-core (SSAC) search in time-dependent road-social networks, where communities are defined by structural cohesiveness, semantic similarity to query keywords, and travel-time proximity to a query location. It introduces two local seed-expansion algorithms, ESSAC (exact) and GSSAC (greedy), which start from the query node and expand outward to identify SSACs, with GSSAC maintaining a single high-quality k-core throughout expansion. Semantic representations are generated by prompting GPT-3.5-Turbo to filter and augment keywords, followed by embedding with text-embedding-3-small to compute keyword similarity metrics $K_c$, while time-dependent cohesiveness uses A*-based travel times $d_{l_v,l_q,t}$ to produce $T_c$. Experimental results on multiple real-world datasets show that GSSAC consistently outperforms baselines in structural, semantic, and time-dependent spatial cohesiveness, while also delivering substantial speedups over ESSAC. The approach offers practical value for tasks such as offline activity planning and detecting coordinated behavior, by enabling accurate, fast, and semantically rich community discovery that respects road-network travel dynamics.

Abstract

Real-world networks often involve both keywords and locations, along with travel time variations between locations due to traffic conditions. However, most existing cohesive subgraph-based community search studies utilize a single attribute, either keywords or locations, to identify communities. They do not simultaneously consider both keywords and locations, which results in low semantic or spatial cohesiveness of the detected communities, and they fail to account for variations in travel time. Additionally, these studies traverse the entire network to build efficient indexes, but the detected community only involves nodes around the query node, leading to the traversal of nodes that are not relevant to the community. Therefore, we propose the problem of discovering semantic-spatial aware k-core, which refers to a k-core with high semantic and time-dependent spatial cohesiveness containing the query node. To address this problem, we propose an exact and a greedy algorithm, both of which gradually expand outward from the query node. They are local methods that only access the local part of the attributed network near the query node rather than the entire network. Moreover, we design a method to calculate the semantic similarity between two keywords using large language models. This method alleviates the disadvantages of keyword-matching methods used in existing community search studies, such as mismatches caused by differently expressed synonyms and the presence of irrelevant words. Experimental results show that the greedy algorithm outperforms baselines in terms of structural, semantic, and time-dependent spatial cohesiveness.

Figures (7)

• Figure 1: A toy example of time-dependent road-social attributed networks. Each node in the social network is associated with keywords and a location on the road network. The travel time (weight) of each road follows a piecewise linear function. For example, the weight $w_{l_{1}, l_{4}} = {(0, 4), (30, 8), (50, 5)}$ for the road $(l_{1}, l_{4})$ is represented by the linear function in the lower right part of the figure, where $(0, 4)$ indicates the travel time between $l_{1}$ and $l_{4}$ is 4 minutes at time 0.
• Figure 2: Illustration of the process for generating semantic vectors using LLMs. Prompts guide GPT-3.5-Turbo to process keywords, and text-embedding-3-small is used to obtain their semantic vectors.
• Figure 3: Illustration of GSSAC. It includes initial community construction, candidate nodes selection, and community expansion.
• Figure 4: Comparison of ESSAC and GSSAC. For simplicity, the abbreviations Fsq_rec and Fsq denote the datasets Foursquare_rec and Foursquare and are used as such in the following figures and tables.
• Figure 5: Results of GSSAC with different $k$.