Crime Hotspot Detection Using K-Means Clustering

Understanding the spatial and temporal patterns that are associated with crime is an asset to urban safety. In this study, k-means clustering, a machine-learning regression algorithm, was used to investigate patterns found in crime data for the city of Chicago. The study goal is to assist public safety agencies in determining how to best use their efforts and resources in responding to crime through clustering similar criminal activity locations. To lend credibility to our analysis, we went through a systematic approach in organizing and cleaning the crime data, such as, normalizing the location and converting text into numbers. The elbow method appears to be the most effective for clustering the city into two clusters, analyzing the crime sites as either high-crime or low-crime locations. Images, including scatterplots and heatmaps, acted as visual aids in displaying the distribution of crime. The results proved the model to be accurate and satisfactory, scoring high on metrics that measure coherence in grouping data. It was also noted that crimes tend to peak on weekends and late-night hours. All in all, this project portrays how we can use machine learning to accentuate policing by offering intelligent ways and techniques for ensuring safety in our cities.