marcelxyz / PreCrimeBot

Python machine learning application for forecasting of London crime

PreCrimeBot

Python application for on-demand prediction of London crime, my graduation project from Queen Mary, University of London.

You can find the relevant project report here.

Installation

This software requires use of Python 3.5+ with Pip. Once these conditions are satisfied, you can install the application into its own virtual environment using the following commands:

python3 -m venv .venv
source .venv/bin/activate
pip install pipenv
pipenv install

This will install all of this project's required dependencies.

Querying

CLI

The simplest and most convenient way to interact with this application using the CLI utility.

After installation and activation of the virtual environment, simply run:

python app/cli.py

This will start an interactive shell environment that is very straightforward to use.

REST API

Optionally you can also make use of the Flask powered REST API. To start it run:

python app/rest_api.py

This will start an HTTP server listening on 127.0.0.1:5000.

The prediction endpoint has the following route:

GET /predict-crime/<date>/<address>

You can then call this endpoint using your favourite HTTP client.

For example, to find predict crime in Westminster for tomorrow, run:

CURL http://127.0.0.1:5000/predict-crime/tomorrow/westminster

Alternatively you can visit http://127.0.0.1:5000/predict-crime/tomorrow/westminster in your browser.

Model retraining

Before retraining the models, you must first download the UK Police data. This process is explained in the data/README.md file.

Once that's done, you can train the models using the training module:

python app/train.py

NOTE: This will attempt to train all the models that are defined in the app/classifier/classifier_factory.py module. If you are only interested in training some of the models, please comment out the irrelevant ones.