Guardians of Tomorrow: Leveraging Responsible AI for Early Detection and Response to Criminal Threats

This archive is distributed in association with the INFORMS Journal on Computing under the MIT License.

The software and data in this repository are a snapshot of the software and data that were used in the research reported on in the paper Guardians of Tomorrow: Leveraging Responsible AI for Early Detection and Response to Criminal Threats by X. Sun, Q. Wang, L. Qiu, and W. Xu.

Cite

To cite the contents of this repository, please cite both the paper and this repo, using their respective DOIs.

https://doi.org/10.1287/ijoc.2023.0488

https://doi.org/10.1287/ijoc.2023.0488.cd

Below is the BibTex for citing this snapshot of the repository.

@misc{ResponsibleAIForCrimeDetection,
  author =        {Xiaotong Sun and Qili Wang and Liangfei Qiu and Wei Xu},
  publisher =     {INFORMS Journal on Computing},
  title =         {Guardians of Tomorrow: Leveraging Responsible AI for Early Detection and Response to Criminal Threats},
  year =          {2025},
  doi =           {10.1287/ijoc.2023.0488.cd},
  url =           {https://github.com/INFORMSJoC/2023.0488},
  note =          {Available for download at https://github.com/INFORMSJoC/2023.0488},
}  

Description

The goal of this repository is to demonstrate our proposed design framework for responsible AI systems that support proactive crime detection and real-time response.

It contains Python codes implementing (1) the data synthesizer used to generate privacy-preserving data and (2) the machine learning models used to achieve crime detection results.

A synthetic dataset is also provided to support replication of the crime detection system presented in our paper.

Requirements

Require Python 3.6.9 or later version.

Install the dependencies using the provided requirements.txt file:

pip install -r requirements.txt

Data Files

The original data for our paper was obtained from a government department and contains private information; therefore, it is confidential.

To support replication, we provide a synthetic dataset in the file /data/sample.csv.

Code Files

There are 2 source code files (in /src).

Data_Synthesizer_for_Preserving_Privacy.py: It includes the construction of the Greedy Bayes network model (Algorithm 1 in DataLearner), the computation of noisy conditional probabilities (Algorithm 2 in DataLearner), the implementation of the DataGenerator, and the analysis of information loss.

Machine_Learning_Methods_for_Predictive_Modeling.py: It includes the training and testing of the machine learning model, as well as the feature importance analysis.

Result Files

There are four types of result files (in /results).

/BNStructure: This folder contains the structure of the learned Greedy Bayes network model.

/AttributeUniqueValues: This folder contains the unique values of each feature for data generation.

/NoisyConditionalProbability: This folder contains the learned noisy conditional probability distributions.

/PrivatePreservingDataset: This folder contains the generated synthetic dataset.