Project Description: K-Means Clustering Analysis of Police Shooting Data

Overview

This project focuses on clustering data into disjoint sets, where data within a set is more similar to each other than to data in other sets. Clustering, a form of unsupervised learning discussed in Chapter 12 of the class reference text, allows me to uncover meaningful patterns in the data without predefined labels. I specifically implement k-means clustering, as outlined in Chapter 12.4 of the text and supported by various instructional videos.

Data Sources

For this project, I analyze two datasets:

1. Locations of Political Violence in India (2016-2024): A dataset detailing incidents of political violence.
2. Locations of Police Shootings in the Continental United States: A dataset detailing incidents of police shootings across the US.

Key Methodologies

• Distance Calculation: Utilizing the Geopy library, I can calculate geodesic distances between latitude and longitude coordinates, essential for clustering analysis.

Issues Addressed

The analysis of police shooting data reveals several critical insights:

1. State Incidents: I examined which states exhibit the highest number of police shooting incidents, comparing results across different cluster configurations (2, 3, 4, 5, and 6 clusters).
2. Age and Death Count Relationship: Investigating the relationship between age, manner of death, and the number of fatalities in the top five states for each of the six clusters.
3. Top States Over Time: Analyzing trends in police shootings in the top five states for each year from 2015 to 2022.
4. Race and City Correlation: Assessing correlations between the top five cities, racial demographics, and the number of police shootings.

Findings

1. Cluster Analysis:

   • In my k-means clustering analysis, Cluster 2 frequently emerged as the largest across various models, with Florida notably represented, indicating significant regional issues. In the six-cluster model, California stood out in Cluster 1.

2. Top States for Incidents:

   • From 2015 to 2021, California consistently recorded the highest number of incidents, followed closely by Texas and Florida, indicating persistent challenges related to police violence.

3. Yearly Trends:

   • California led the incidents each year, with Texas and Florida regularly appearing among the top three states, highlighting ongoing policing issues.

4. Racial Disparities:

   • The analysis indicated that police shooting incidents disproportionately affect Black individuals, particularly in cities like Los Angeles, Phoenix, and Houston, raising serious concerns about systemic discrimination in law enforcement.

5. Urban Concentration:

   • Most police shooting incidents were found concentrated in urban areas, suggesting unique challenges in policing in these densely populated regions.

Appendix A: Method

1. Data Collection:

   • The dataset was sourced from the specified class link, focusing on the 'fatal-police-shootings-data-continental-US.csv' file.

2. Data Preparation:

   • A thorough examination of the dataset was conducted to ensure completeness and accuracy, documenting all steps for reproducibility.

3. Variable Creation:

   • Key columns for correlation analysis included age, race, state, year, manner_of_death, and gender.

4. Libraries Used:

   • Pandas: For data manipulation and handling missing values.
   • Folium: For creating interactive maps and visualizing data points.
   • Scikit-learn: For applying the K-means clustering algorithm.
   • Seaborn: For generating a color palette for clusters.

5. Analytic Method:

   • Data preprocessing, legend addition for clarity in maps, descriptive statistics to understand incident distribution, application of the K-means algorithm for clustering, and visualization of clustered data to identify spatial patterns across the United States. """