Task: Crime Forecasting

Task: Crime Forecasting#

Overview#

The task involves leveraging data science techniques to participate in a real-time crime forecasting challenge inspired by the National Institute of Justice’s Real-Time Crime Forecasting Challenge. Your goal is to analyze and forecast place-based crime patterns using provided data and additional relevant sources.

Objective#

Develop an algorithm or model that accurately predicts crime hotspots in a specified jurisdiction using the provided Portland Police Bureau (PPB) Calls-for-Service (CFS) dataset. Predictions will focus on specific crime categories over various timeframes.

Task Deliverables#

  1. Data Analysis and Preprocessing

    • Load and explore the CFS dataset to identify patterns, trends, and features.

    • Handle any missing data or inconsistencies in the dataset.

  2. Model Development

    • Develop a machine learning or statistical model for forecasting crime locations.

    • Focus on All Calls-for-Service

    • Forecast crime hotspots for time periods of:

      • Two weeks

  3. Evaluation Metrics

    • Use the Prediction Accuracy Index (PAI) to measure how well your model predicts hotspots.

    • Use the Prediction Efficiency Index (PEI) to evaluate the efficiency of your predictions.

    • Use the March-May, 2017 Calls-for-Service Data as the test set to evaluate your solution’s PAI and PEI.

  4. Visualization

    • Provide visualizations of the forecasted crime hotspots.

    • Clearly indicate predicted high-risk areas.

  5. Report

    • Submit a concise report detailing:

      • Data exploration and preprocessing steps.

      • Model development and chosen features.

      • Evaluation of your model’s performance.

      • Discussion of challenges faced and potential improvements.

    • Include your visualizations in the report.

  6. Submission

    • Submit your code as a GitHub repository.

    • Share the GitHub repository link with me.

Data Access#

Download the provided data set from the challenge webpage. Additional data sources may be used to enrich your analysis, but ensure they are appropriately cited.

Submission Format#

  • Share your GitHub repository containing:

    • Python code used for analysis and forecasting.

    • Forecast results for each timeframe in a .csv format.

    • Visualizations in .png or .jpg format.

    • A report in PDF format.

    • Your resume

    • One page statement to descript your … for the GA positions.

Evaluation Criteria#

  1. Prediction Accuracy and Efficiency: How accurate and efficient your forecasts are based on PAI and PEI metrics.

  2. Innovation: Use of novel techniques or approaches in your analysis and forecasting.

  3. Clarity of Visualizations and Report: How well your findings are presented and explained.

Deadline#

Submit your completed task by 05/12/2025.

Resources#