Skip to content

Latest commit

 

History

History

States-Grouping_Assault-Prediction

README.md

US States Grouping and Assault Rate Prediction

Date: 2024-09-20

Introduction

Brief Overview:
This project analyses socio-economic and crime data from various datasets to group US states based on their characteristics and identify factors that significantly predict assault rates. It was undertaken as part of a technical assessment for a job application at Singapore public agency.

The goal was to demonstrate analytical and statistical skills, including clustering and regression analysis, in the context of real-world data challenges. Insights from this project can aid in understanding socio-economic patterns and crime trends across the US.

Objectives:

  • Group US states based on socio-economic and crime characteristics using clustering analysis.
  • Identify significant predictors of assault rates through regression models.
  • Provide data-driven insights to better understand socio-economic disparities and crime rates.

Table of Contents

  1. Introduction
  2. Installation
  3. Usage
  4. Data
  5. Methodology
  6. Results
  7. Project Structure
  8. License
  9. Contact Information

Installation

Prerequisites:

  • R version 4.0+
  • RStudio

Dependencies: Install the required R libraries by running the following in R:

install.packages(c("tidyverse", "cluster", "factoextra", "broom", "car", "ggcorrplot", "usmap"))

Usage

Instructions:

  1. Access States-Grouping_Assault-Prediction.md to view the pre-run analysis.

    or

  2. Download the folder, States_Grouping_Assault_Prediction.

  3. Open States_Grouping_Assault_Prediction.Rmd in RStudio.

  4. Knit the file to generate an HTML report. Alternatively, run the code chunks sequentially to reproduce the analysis.

Data

Data Sources:

  • USArrest.csv: Contains crime data, including assault rates, for US states.
  • USstatex77.csv: Includes socio-economic indicators for US states.

Data Description:

  • USArrest.csv: Columns include Murder, Assault, UrbanPop, and Rape.
  • USstatex77.csv: Contains indicators like Population, Income, Illiteracy, and Life Expectancy.

Data Processing:

  • Merged multiple datasets on the State column.
  • Normalised numeric variables for clustering.
  • Handled missing data and multicollinearity issues for regression analysis.

Methodology

Techniques Used:

  • Clustering Analysis: Grouped states using K-means clustering with 4 clusters.
  • Regression Analysis: Identified significant predictors of assault rates using multiple linear regression.

Tools:

  • Programming Languages: R
  • Libraries: tidyverse, cluster, factoextra, ggcorrplot, car, usmap

Results

Findings:

  • Cluster Profiles: States were grouped into clusters based on crime rates, income, urbanization, and education.
  • Regression Insights: Higher urban population, higher rape rates, and lower life expectancy were associated with higher assault rates.

Visualisations:

  • Cluster Visualization: Principal Component Analysis (PCA) plot showing clusters.
  • Correlation Matrix: Displays relationships among variables.

Interpretation:

  • Policies targeting urban population management and crime reduction can potentially mitigate assault rates.

Project Structure

Directory Tree:

States-Grouping_Assault-Prediction/
│
├── README.md                                              # Project overview and instructions
├── States-Grouping_Assault-Prediction.md                  # Main analysis report
├── States-Grouping_Assault-Prediction.Rmd                 # Main analysis script
├── US_Arrest_Data/                                        # Data files (e.g., USArrest.csv, USstatex77.csv)
└── States-Grouping_Assault-Prediction_files/figure-gfm/   # Visuals used in the report

Key Files:

  • States-Grouping_Assault-Prediction.md: The main Markdown report with analysis, viewable directly on GitHub.
  • States-Grouping_Assault-Prediction.Rmd: The main R Markdown file containing the analysis.
  • US_Arrest_Data/: Folder containing the input datasets.
  • States-Grouping_Assault-Prediction_files/figure-gfm/: Folder to store generated outputs (e.g., reports, tables).

License

This project is intended for submission as part of a technical assessment. The content and code are not intended for public distribution, reproduction, or commercial use without explicit permission from the Author.

Contact Information

Author: Ou Yang Yu