This week, I applied Principal Component Analysis (PCA) and K-means clustering to explore patterns of protests, riots, and violence across U.S. states.

PCA reduced the dataset to two main dimensions (PC1 and PC2), capturing over 96% of the total variation. This made it easier to compare states based on their normalized event rates per million people.

Using the PCA results, I grouped states into three clusters. Most states clustered tightly together, while a few outliers formed their own distinct clusters.

Noticeably:

• Cluster 0 (Oregon) and Cluster 1 (District of Columbia) were positioned far outside the main group in the PCA plot.
• District of Columbia (Cluster 1) had the highest protest rate by far, making it a clear outlier.
• Oregon (Cluster 0) stood out with high levels of both protests and riots, reflected in its extreme PC2 value.
• The rest of the states (Cluster 2) shared more similar patterns and grouped near the origin.

The top 5 states with the highest PC1 values (indicating highest overall event intensity) were:

State	PC1	PC2	Cluster
District of Columbia	10.76	-0.69	1
Oregon	2.45	3.51	0
Vermont	1.16	-0.82	2
New Mexico	0.77	-0.77	2
New York	0.64	0.44	2