Crime Data Clustering

This project performs clustering analysis on U.S. crime data using Hierarchical, K-Means, and DBSCAN algorithms. The aim is to identify the optimal number of clusters and derive meaningful inferences about crime patterns based on features like Murder, Assault, Urban Population, and Rape.

Table of Contents

• Data Description
• Clustering Techniques
  • Hierarchical Clustering
  • K-Means Clustering
  • DBSCAN Clustering
• Visualizations
• Inferences
• Setup and Usage
• Results

---

Data Description

• Murder: Murder rates in different places of the United States.
• Assault: Assault rates in different places of the United States.
• UrbanPop: Urban population in different places of the United States.
• Rape: Rape rates in different places of the United States.

Clustering Techniques

Hierarchical Clustering

• Used dendrograms to identify the number of clusters.
• Applied Agglomerative Clustering to group similar regions.

K-Means Clustering

• Used Elbow Curve to find the optimal number of clusters.
• Clustered data using K-Means and analyzed intra-cluster similarities.

DBSCAN Clustering

• Applied Density-Based Spatial Clustering for identifying arbitrarily shaped clusters.
• Tuned eps and min_samples parameters for optimal clustering.

Visualizations

• Dendrograms for hierarchical clustering.
• Elbow curve for K-Means optimization.
• Scatter plots for cluster visualization.

Inferences

• Hierarchical and K-Means clustering identified consistent cluster patterns.
• DBSCAN was sensitive to parameter tuning but struggled with noise in the data (silhouette score: -0.268).
• Urban population and crime rates influence clustering significantly.

Setup and Usage

1. Clone the repository:

   git clone https://github.com/R-Mahesh45/crime-data-clustering.git

2. Install required libraries:

   pip install pandas numpy matplotlib seaborn scipy scikit-learn

3. Run the clustering analysis:

   python clustering_analysis.py

Results

• Hierarchical Clustering: X clusters were formed.
• K-Means Clustering: Y clusters were optimal based on the elbow method.
• DBSCAN Clustering: Challenging to form clusters due to noise.