Project 2 – Week 2

After cleaning the United States conflict data from ACLED data, my question is: What groups of states share similar patterns of political protests and violence?

To identify meaningful patterns and groupings among states, I will employ multiple analytical techniques:

1. Clustering analysis
2. K-means algorithm
3. Principal Component Analysis (PCA)

 I created a comprehensive dataset that counts various political event types across all states. This dataset tracks six distinct categories:

• Battles
• Explosions/Remote violence
• Protests
• Riots
• Strategic developments
• Violence against civilians

Here’s what the data looks like for the first 10 states:

States	Battles	Explosions/ Remote violence	Protests	Riots	Strategic developments	Violence against civilians
Alabama	0	0	622	4	60	5
Alaska	0	0	143	0	3	0
Arizona	1	0	719	12	479	11
Arkansas	0	1	332	3	27	4
California	1	8	8543	187	332	48
Colorado	1	0	868	13	78	9
Connecticut	0	0	1108	10	53	1
Delaware	0	0	162	1	6	1
District of Columbia	1	0	1502	36	79	10
Florida	0	0	3003	29	144	25

Descriptive statistics:

	Battles	Explosions/ Remote violence	Protests	Riots	Strategic developments	Violence against civilians
Standard deviation	0.7736	1.1944	1551.7963	44.1649	82.5585	8.3162
Skewness	2.1998	5.6165	3.4210	3.4518	3.2169	2.9503
Kurtosis	4.2542	35.2769	13.3489	11.3228	13.0840	11.8533

Scatterplot matrix :

This scatterplot matrix reveals relationships between different political event types across US states. Key observations include:

• Most states have few battles and explosions, with only rare exceptions.
• Protests are the most common event type, showing wide variation across states.
• States with more protests often have more riots as well.
• Most states experience low levels of strategic developments and violence against civilians, though some notable outliers exist.

This visualization helps identify potential patterns and outliers before our clustering analysis.