Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CrimeAlarm: Towards Intensive Intent Dynamics in Fine-grained Crime Prediction

Kaixi Hu, Lin Li, Qing Xie, Xiaohui Tao, Guandong Xu

TL;DR

A fine-grained sequential crime prediction framework, CrimeAlarm, that equips with a novel mutual distillation strategy inspired by curriculum learning that outperforms state-of-the-art methods in terms of NDCG@5 and accuracy measures.

Abstract

Granularity and accuracy are two crucial factors for crime event prediction. Within fine-grained event classification, multiple criminal intents may alternately exhibit in preceding sequential events, and progress differently in next. Such intensive intent dynamics makes training models hard to capture unobserved intents, and thus leads to sub-optimal generalization performance, especially in the intertwining of numerous potential events. To capture comprehensive criminal intents, this paper proposes a fine-grained sequential crime prediction framework, CrimeAlarm, that equips with a novel mutual distillation strategy inspired by curriculum learning. During the early training phase, spot-shared criminal intents are captured through high-confidence sequence samples. In the later phase, spot-specific intents are gradually learned by increasing the contribution of low-confidence sequences. Meanwhile, the output probability distributions are reciprocally learned between prediction networks to model unobserved criminal intents. Extensive experiments show that CrimeAlarm outperforms state-of-the-art methods in terms of NDCG@5, with improvements of 4.51% for the NYC16 and 7.73% for the CHI18 in accuracy measures.

CrimeAlarm: Towards Intensive Intent Dynamics in Fine-grained Crime Prediction

TL;DR

A fine-grained sequential crime prediction framework, CrimeAlarm, that equips with a novel mutual distillation strategy inspired by curriculum learning that outperforms state-of-the-art methods in terms of NDCG@5 and accuracy measures.

Abstract

Granularity and accuracy are two crucial factors for crime event prediction. Within fine-grained event classification, multiple criminal intents may alternately exhibit in preceding sequential events, and progress differently in next. Such intensive intent dynamics makes training models hard to capture unobserved intents, and thus leads to sub-optimal generalization performance, especially in the intertwining of numerous potential events. To capture comprehensive criminal intents, this paper proposes a fine-grained sequential crime prediction framework, CrimeAlarm, that equips with a novel mutual distillation strategy inspired by curriculum learning. During the early training phase, spot-shared criminal intents are captured through high-confidence sequence samples. In the later phase, spot-specific intents are gradually learned by increasing the contribution of low-confidence sequences. Meanwhile, the output probability distributions are reciprocally learned between prediction networks to model unobserved criminal intents. Extensive experiments show that CrimeAlarm outperforms state-of-the-art methods in terms of NDCG@5, with improvements of 4.51% for the NYC16 and 7.73% for the CHI18 in accuracy measures.
Paper Structure (36 sections, 12 equations, 7 figures, 3 tables, 1 algorithm)

This paper contains 36 sections, 12 equations, 7 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Contributions.
  3. Preliminaries
  4. Problem Statement
  5. Fined-grained Sequential Crime Prediction.
  6. Decoupled Knowledge Distillation
  7. Why knowledge distillation works?
  8. CrimeAlarm
  9. Overview
  10. Simple-to-Difficult Target Crime Distillation
  11. Analysis about Target.
  12. Difficult Target Crime Distillation.
  13. Target Crime Distillation Loss.
  14. Less-to-More Non-target Crime Distillation
  15. Analysis about Non-targets.
  16. ...and 21 more sections

Figures (7)

  • Figure 1: An illustration of intent dynamics behind crime events. Criminal intents are grouped into either property or violence. It is observed that both spot A and spot B contain multiple criminal intents (as flagged by different colors), and these intents switch frequently and/or progress differently.
  • Figure 2: The framework of our proposed CrimeAlarm. In the early distillation phase, spot-shared intents are learned via simple target, and less yet frequent non-target crime events. In the later phase, spot-specific intents are captured via difficult target and more non-target crime events.
  • Figure 3: Ablation study.
  • Figure 4: Growing peers.
  • Figure 5: Discussion on densities.
  • ...and 2 more figures