Multi-Type Urban Crime Prediction

Authors: Xiangyu Zhao, Wenqi Fan, Hui Liu, Jiliang Tang4388-4396

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on real-world datasets validate the effectiveness of our framework. In this section, we conduct extensive experiments to evaluate the effectiveness of the proposed framework.
Researcher Affiliation Academia Xiangyu Zhao1, Wenqi Fan2 , Hui Liu3, Jiliang Tang3 1City University of Hong Kong, 2The Hong Kong Polytechnic University, 3Michigan State University
Pseudocode No The paper describes optimization steps using mathematical equations, but it does not present them as a formally labeled 'Pseudocode' or 'Algorithm' block.
Open Source Code No The paper does not provide any concrete access to source code for the methodology described.
Open Datasets No The data of K=7 types of crime is collected from 07/01/2012 to 06/30/2013 (T=365 days) in New York City. We respectively segment the city into disjointed 2km 2km grids (regions), and select N=100 regions with the most reported crimes. For the feature matrices, we collect multiple data resources that are related to crime: historical crime, stop-and-frisk, weather, Point of Interests (region function), human mobility and 311 public-service complaint data. No specific links or formal citations for public access are provided.
Dataset Splits No The paper mentions training and test samples, and uses cross-validation for parameter selection, but does not specify a distinct validation dataset split with proportions or counts for reproduction.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, or memory) used for running experiments.
Software Dependencies No The paper mentions using the ADMM technique and references related work, but it does not specify any software names with version numbers (e.g., Python version, specific libraries with versions).
Experiment Setup Yes We select parameters of the proposed framework such as α, β, γ, ρ and σ by cross-validation; (ii) For a fair comparison, we conduct parameter-tuning for baselines. More details about parameter selection will be discussed in following subsections. CCC achieves the best performance when α = 2 for 1-day prediction, while α = 3 for 7-day prediction. The performance achieves the peak when β = 1.25 for 1-day prediction and β = 1 for 7-day prediction and CCC approaches the best performance when γ = 0.5 for both 1-day and 7-day prediction.