Comparative Study of Clustering Algorithms: Scenario Based on Boston Crime Dataset

In the study of crime analysis, using clustering techniques to understand patterns in crime-prone areas is crucial. This research paper focuses on grouping regions with higher crime rates using various clustering methods and then thoroughly compares and evaluates the results. We have applied six different clustering methods: K-Means, K-Prototypes, Agglomerative Hierarchical Clustering, DBSCAN Clustering, Fuzzy C-Means clustering, and Gaussian Mixture Models (GMMs) to the Boston City Crime Dataset. These methods were carefully examined to understand their performance. For example, K-Means and K-Prototypes create clusters, Agglomerative Clustering organizes data hierarchically, Fuzzy C-Means extends the principles of K-Means for softer clustering, DBSCAN uses a unique approach based on graph density, and GMM effectively handles complex data structures while allowing for probabilistic assignments.This research enhances the field by introducing inventive algorithmic applications, systematically assessing their effectiveness, rigorously evaluating their performance, and providing practical insights into their implementation for analyzing crime data on extensive datasets, along with aiming to demonstrate the effectiveness of different clustering techniques in identifying areas or clusters of higher crime rates.