farihat13 / STHSL

Code for STHSL model proposed in Spatial-Temporal Hypergraph Self-Supervised Learning for Urban Crime Prediction in ICDE'2022

STHSL

A pytorch implementation for the paper:
Spatial-Temporal Hypergraph Self-Supervised Learning for Crime Prediction
In ICDE 2022.

Introduction

Spatial-Temporal Hypergraph Self-Supervised Learning for Crime Prediction (STHSL) is a spatio-temporal prediction networks. By adding self-supervised learning methods as auxiliary tasks, STHSL can tackle the label scarcity issue in crime prediction.

Structure

• Datasets: including NYC and CHI datasets used in our experiments, which are released by and available at STSHN
• Save: model save path，for testing
• model: model of STHSL
• others: files required for model training

Environment requirement

The code can be run in the following environments, other version of required packages may also work.

• python==3.9.7
• numpy==1.22.3
• pytorch==1.9.0

Run the codes

• NYC-Crimes dataset: Train and Test

python train.py --data NYC

python test.py --data NYC --checkpoint ./Save/NYC/your_file_names

• Chicago-Crimes dataset: Train and Test

python train.py --data CHI

python test.py --data CHI --checkpoint ./Save/CHI/your_file_names