San Francisco Crime Data

Collaborators:

Name	github.com
Betty Zhou	@bettybhzhou
Ian Flores	@ian-flores

Project Overview

Objective

The goal of this project is to implement a classification of the San Francisco crime data with a decision tree to predict the resolution of a crime instance. The resolution of a crime instance can be either "processed" or "non-processed". Processed indicates that a crime instance resulted in a subject being processed into the justice system. The following table outlines crime resolutions that were classified as processed or non-processed:

Non-Processed	Processed
NONE	ARREST, BOOKED
CLEARED-CONTACT JUVENILE FOR MORE INFO	ARREST, CITED
UNFOUNDED	NOT PROSECUTED
JUVENILE ADMONISHED	PSYCHOPATHIC CASE
EXCEPTIONAL CLEARANCE	JUVENILE CITED
JUVENILE DIVERTED	JUVENILE BOOKED
	LOCATED
	PROSECUTED BY OUTSIDE AGENCY
	COMPLAINANT REFUSES TO PROSECUTE
	DISTRICT ATTORNEY REFUSES TO PROSECUTE
	PROSECUTED FOR LESSER OFFENSE

This project will address the following predictive question:
What are the strongest predictors for whether a crime instance in San Francisco resulted in someone being "processed" or "not processed" into the justice system?

Dataset

The dataset contains San Francisco crime data from 2003 to May 2018 with each crime instance having the following features: incident ID, category of crime, description of crime, day of week of incident, date of incident, time of incident, police district, resolution of incident, address, longitude, latitude and location of incident.

The dataset can be found at the following link:

• Link to San Francisco crime data website

The dataset can be loaded using this R script or this Python script. The following is a preview of the dataset in R:

Analysis and Results

The final report of the project can be found here

Analysis Pipeline

Usage using Docker:

To run this analysis using Docker, execute the following steps:

1. Clone this repository

2. Pull the Docker image from Docker Hub using the follwing command:

   docker pull bettybhz/san_francisco_crime_resolution_model
   

3. Navigate to the root of this project on your computer using the command line.

4. Execute the following command by filling in <PATH_ON_YOUR_COMPUTER> with the absolute path to the root of this project on your computer).

   docker run --rm -v PATH_ON_YOUR_COMPUTER:/home/San_Francisco_Crime_Resolution_Model bettybhz/san_francisco_crime_resolution_model make -C 'home/San_Francisco_Crime_Resolution_Model' all
   

   Execute the following command to clean up the analysis:

   docker run --rm -v PATH_ON_YOUR_COMPUTER:/home/San_Francisco_Crime_Resolution_Model bettybhz/san_francisco_crime_resolution_model make -C 'home/San_Francisco_Crime_Resolution_Model' clean
   
   

Usage from command line:

1. Clone this repository

2. From the root of the project, run the following commands:

python src/01_clean_data.py 1000 data/san_francisco_clean.csv
python src/02_EDA.py data/san_francisco_clean.csv results/figures/
python src/03_feature_engineering.py data/san_francisco_clean.csv data/san_francisco_features.csv
python src/04_decison_tree.py data/san_francisco_features.csv data/feature_results.csv
Rscript src/03_Exploratory_SF_map.R data/san_francisco_clean.csv results/figures/
Rscript -e "rmarkdown::render('docs/san_francisco_report.Rmd')"

Or, run all scripts from the root using the following command:

make all

To clean up analysis, run:

make clean

Dependencies:

• R (v 3.5.1) & R libraries

  • rmarkdown v1.10
  • knitr v1.20
  • tidyversev 1.2.1
  • ggmap v2.6.1
  • maps v3.3.0

• Python (v 3.6.5) & Python libraries:

  • matplotlib v2.2.2
  • numpy v1.14.3
  • seaborn v0.9.0
  • pandas v.23.0
  • argparse v1.1
  • sklearn v0.19.1
  • pendulum v2.0.4