Analysis on Baltimore PD Crime Data

Capstone project for the spring of 2020. Completed on April 28th, 2020. Please see the FinalReport.pdf or our dashboard for more details on the analysis. A non-technical summary can also be found on the dashboard.

Project Objectives

Our primary goal is to find the best way to convey Baltimore’s crime data in a way that is helpful to the Baltimore Police Department. Our secondary goals were:

1. To predict what type of crime occured (violent or nonviolent)
2. To predict how many crimes will occur in a given day
3. To create visualizations that present interesting, informative findings

Authors

1. Caroline Simmons
2. Sejal Patrikar
3. Jeannette Jiang
4. Jenny Jang
   

Special thanks to our advisor Jeffery Holt for helping us throughout this project.

Data

Raw Data Files

• BPD_Part_1_Victim_Based_Crime_Data_April25_2020.csv: raw data from the Baltimore Police Department found here. Each row is an offense, with variables such as location, time, date, offense description, weapons, ect. This source updates weekly, and was last updated on April 25th, 2020.
• unemployment_version_final.csv : Baltimore’s monthly unemployment rates from the Maryland Department of Labor from 2015-2019. Minor data cleaning has been done in Excel.
• weather_final.csv : daily weather data from 2015-2019 from the National Center for Environmental Information containing data on precipitation, snow, snow depth, average temperature, etc. Minor data cleaning has been done in Excel.

Cleaned/Created Data Files

• baltimore_daily_cleaned.csv: This contains all variables from 2014-2019 summarized by date. Run raw BPD data through DataCleaning_part1.R and then through DataCleaning_part2.ipynb to get this file.
• baltimore_cleaned.csv: This contains all variables from 2014-2019 as a transaction table. Run raw BPD data through DataCleaning_part1.R and then through DataCleaning_part2.ipynb to get this file.
• neighborhood_balt.csv: created data file compiling location and rank of top 5 most dangerous neighborhoods in 2020 for Baltimore.

Predicting Daily Crime Totals

• TreeMethod.R contains several tree models
• TimeSeries.R contains 2 SARIMA models
• LinearRegression.R contains a simple MLR model.

Predicting Crime Type

• LogisticRegression.R - contains 2 logistic regression models.

Visualizations

• TimeSeries.R - contains COVID-19 time series plot
• DataVisualization.Rmd - contains code for several graphs

Dashboard

• Capstone_dashboard.rmd- contains the code to run the flexdashboard BaltimoreCrime.
  NOTE: ggmap package is used in Capstone_dashboard.rmd. To run that file, one needs to register their device with an API in order to use ggmap and produce certain visualizations that require ggmap software.