Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Skyline-based exploration of temporal property graphs

Evangelia Tsoukanara, Georgia Koloniari, Evaggelia Pitoura

TL;DR

The paper addresses detecting significant temporal events in evolving temporal property graphs by introducing a parameter-free skyline-based exploration over aggregated graphs. It formalizes a temporal property graph model with time-varying and static properties, defines strict and loose semantic operators for evolution and aggregation, and develops both individual and unified evolution skylines to capture trade-offs between interval length and event counts. A top-$k$ unified skyline extends this with domination-degree ranking, and the approach is implemented atop GraphTempo with efficient incremental computation. Empirical evaluation on DBLP, MovieLens, and Primary School datasets demonstrates the method's efficiency and its ability to reveal interpretable, high-signal temporal patterns, including how aggregation and semantic choices affect skyline size and discovery. Overall, the work provides a robust, scalable framework for discovering meaningful temporal patterns in complex graphs without parameter tuning, with potential impact on analytics in social, bibliographic, and contact-network domains.

Abstract

In this paper, we focus on temporal property graphs, that is, property graphs whose labeled nodes and edges as well as the values of the properties associated with them may change with time. For instance, consider a bibliographic network, with nodes representing authors and conferences with properties such as gender and location respectively, and edges representing collaboration between authors and publications in conferences. A key challenge in studying temporal graphs lies in detecting interesting events in their evolution, defined as time intervals of significant stability, growth, or shrinkage. To address this challenge, we build aggregated graphs, where nodes are grouped based on the values of their properties, and seek events at the aggregated level, for example, time intervals of significant growth in the collaborations between authors of the same gender. To locate such events, we propose a novel approach based on unified evolution skylines. A unified evolution skyline assesses the significance of an event in conjunction with the duration of the interval in which the event occurs. Significance is measured by a set of counts, where each count refers to the number of graph elements that remain stable, are created, or deleted, for a specific property value. For example, for property gender, we measure the number of female-female, female-male, and male-male collaborations. Lastly, we share experimental findings that highlight the efficiency and effectiveness of our approach.

Skyline-based exploration of temporal property graphs

TL;DR

The paper addresses detecting significant temporal events in evolving temporal property graphs by introducing a parameter-free skyline-based exploration over aggregated graphs. It formalizes a temporal property graph model with time-varying and static properties, defines strict and loose semantic operators for evolution and aggregation, and develops both individual and unified evolution skylines to capture trade-offs between interval length and event counts. A top- unified skyline extends this with domination-degree ranking, and the approach is implemented atop GraphTempo with efficient incremental computation. Empirical evaluation on DBLP, MovieLens, and Primary School datasets demonstrates the method's efficiency and its ability to reveal interpretable, high-signal temporal patterns, including how aggregation and semantic choices affect skyline size and discovery. Overall, the work provides a robust, scalable framework for discovering meaningful temporal patterns in complex graphs without parameter tuning, with potential impact on analytics in social, bibliographic, and contact-network domains.

Abstract

In this paper, we focus on temporal property graphs, that is, property graphs whose labeled nodes and edges as well as the values of the properties associated with them may change with time. For instance, consider a bibliographic network, with nodes representing authors and conferences with properties such as gender and location respectively, and edges representing collaboration between authors and publications in conferences. A key challenge in studying temporal graphs lies in detecting interesting events in their evolution, defined as time intervals of significant stability, growth, or shrinkage. To address this challenge, we build aggregated graphs, where nodes are grouped based on the values of their properties, and seek events at the aggregated level, for example, time intervals of significant growth in the collaborations between authors of the same gender. To locate such events, we propose a novel approach based on unified evolution skylines. A unified evolution skyline assesses the significance of an event in conjunction with the duration of the interval in which the event occurs. Significance is measured by a set of counts, where each count refers to the number of graph elements that remain stable, are created, or deleted, for a specific property value. For example, for property gender, we measure the number of female-female, female-male, and male-male collaborations. Lastly, we share experimental findings that highlight the efficiency and effectiveness of our approach.
Paper Structure (16 sections, 6 theorems, 12 figures, 5 tables, 1 algorithm)

This paper contains 16 sections, 6 theorems, 12 figures, 5 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. The temporal property graph model
  3. Temporal graph operators
  4. Graph evolution and aggregation
  5. Graph exploration
  6. Evolution skyline
  7. Skyline-based exploration
  8. Experimental evaluation
  9. Performance evaluation
  10. Qualitative evaluation
  11. Skyline size.
  12. Individual skylines for stability.
  13. Unified skylines for locating important time periods.
  14. Comparing individual and unified skylines.
  15. Related work
  16. ...and 1 more sections

Key Result

Proposition 1

Given an aggregate event graph $G_{\gamma}[(T_2, T_1), P]$, if $p$ and $p'$ correspond to values of properties in $P$, then $count(G_{\gamma}[(T_2, T_1)], l_e, p, p')$ is equal to the weight property of the edge labeled $l_e$ between the nodes with labels $p$ and $p'$ in $G_{\gamma}[(T_2, T_1), P]$.

Figures (12)

  • Figure 1: An example of a temporal property graph for a bibliographic network defined on $T = [1, 3]$.
  • Figure 2: $G[T_1 \sqcup T_2]$ with (a) loose and (b) strict semantics.
  • Figure 3: Graph instance on time point $3$ for the graph of Fig. \ref{['fig:tpg']}.
  • Figure 4: Graph $G[[3], \{gender\}]$ aggregated by author label and gender property for the graph of Fig. \ref{['fig:t3']}.
  • Figure 5: Graph $G[[3], \{gender, topic\}]$ aggregated by author, conference labels and gender, topic properties for the graph of Fig. \ref{['fig:t3']}.
  • ...and 7 more figures

Theorems & Definitions (16)

  • definition thmcounterdefinition: Temporal Property Graph
  • definition thmcounterdefinition: Graph aggregation
  • Proposition 1
  • definition thmcounterdefinition: Individual evolution skyline
  • lemma thmcounterlemma
  • proof
  • definition thmcounterdefinition: Unified evolution skyline
  • lemma thmcounterlemma
  • proof
  • lemma thmcounterlemma
  • ...and 6 more