Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An Event-centric Framework for Predicting Crime Hotspots with Flexible Time Intervals

Jiahui Jin, Yi Hong, Guandong Xu, Jinghui Zhang, Jun Tang, Hancheng Wang

TL;DR

A novel event-centric framework that incorporates a continuous-time attention network to capture correlations between crime events, which learns crime context features, representing general crime patterns across time points and locations, and a type-aware spatiotemporal point process that learns crime-evolving features.

Abstract

Predicting crime hotspots in a city is a complex and critical task with significant societal implications. Numerous spatiotemporal correlations and irregularities pose substantial challenges to this endeavor. Existing methods commonly employ fixed-time granularities and sequence prediction models. However, determining appropriate time granularities is difficult, leading to inaccurate predictions for specific time windows. For example, users might ask: What are the crime hotspots during 12:00-20:00? To address this issue, we introduce FlexiCrime, a novel event-centric framework for predicting crime hotspots with flexible time intervals. FlexiCrime incorporates a continuous-time attention network to capture correlations between crime events, which learns crime context features, representing general crime patterns across time points and locations. Furthermore, we introduce a type-aware spatiotemporal point process that learns crime-evolving features, measuring the risk of specific crime types at a given time and location by considering the frequency of past crime events. The crime context and evolving features together allow us to predict whether an urban area is a crime hotspot given a future time interval. To evaluate FlexiCrime's effectiveness, we conducted experiments using real-world datasets from two cities, covering twelve crime types. The results show that our model outperforms baseline techniques in predicting crime hotspots over flexible time intervals.

An Event-centric Framework for Predicting Crime Hotspots with Flexible Time Intervals

TL;DR

A novel event-centric framework that incorporates a continuous-time attention network to capture correlations between crime events, which learns crime context features, representing general crime patterns across time points and locations, and a type-aware spatiotemporal point process that learns crime-evolving features.

Abstract

Predicting crime hotspots in a city is a complex and critical task with significant societal implications. Numerous spatiotemporal correlations and irregularities pose substantial challenges to this endeavor. Existing methods commonly employ fixed-time granularities and sequence prediction models. However, determining appropriate time granularities is difficult, leading to inaccurate predictions for specific time windows. For example, users might ask: What are the crime hotspots during 12:00-20:00? To address this issue, we introduce FlexiCrime, a novel event-centric framework for predicting crime hotspots with flexible time intervals. FlexiCrime incorporates a continuous-time attention network to capture correlations between crime events, which learns crime context features, representing general crime patterns across time points and locations. Furthermore, we introduce a type-aware spatiotemporal point process that learns crime-evolving features, measuring the risk of specific crime types at a given time and location by considering the frequency of past crime events. The crime context and evolving features together allow us to predict whether an urban area is a crime hotspot given a future time interval. To evaluate FlexiCrime's effectiveness, we conducted experiments using real-world datasets from two cities, covering twelve crime types. The results show that our model outperforms baseline techniques in predicting crime hotspots over flexible time intervals.

Paper Structure

This paper contains 27 sections, 21 equations, 12 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Deep Crime Prediction
  4. Spatio-temporal Point Process
  5. Continuous-time Sequence Models
  6. Preliminaries
  7. FlexiCrime: An Event-centric Framework
  8. Overview of Framework
  9. Extracting Crime Context Feature with Continuous-time Attention Network
  10. Target-aware Crime Event Encoder
  11. Crime Context Feature Extraction
  12. Extracting Crime Evolving Feature with Type-aware Spatiotemporal Point Process
  13. Crime Risk Intensity
  14. Crime Evolving Feature Extraction
  15. Flexible Interval Prediction
  16. ...and 12 more sections

Figures (12)

  • Figure 1: Comparison of true and predicted crime hotspots in New York City on Jan 31, 2018. The crime hotspots across a city appear significantly different during various time intervals (TI). Existing methods usually predicted crime hotspots with specified time granularities (day or half-day), which can result in inaccurate predictions if the predefined time intervals used during training do not align with the target time interval during prediction.
  • Figure 2: Overview of FlexiCrime
  • Figure 3: Continuous-time attention network
  • Figure 4: Type-aware spatiotemporal point process
  • Figure 5: Flexible-interval prediction
  • ...and 7 more figures

Theorems & Definitions (1)

  • Example 1