Unsupervised clustering-based-identification of high-risk urban zones for women’s safety using geospatial data

In many Indian cities, ensuring safety for women continues to be a pressing concern due to factors such as inadequate lighting, low public presence, and limited access to emergency infrastructure. Traditional methods often lack spatial specificity and real-time adaptability. To address this gap, the present study proposes a geospatial data-driven framework for identifying high-risk urban zones using unsupervised machine learning techniques. A composite safety index was developed by assigning weights to key indicators—street lighting density, pedestrian visibility, proximity to transport nodes, and crowd presence based on publicly available demographic and infrastructure datasets. Using this index, two clustering algorithms, K-Means and DBSCAN, were applied to classify urban regions based on relative safety. DBSCAN outperformed K-Means by effectively detecting 18 spatially irregular and isolated unsafe zones, especially in areas with sparse population data. While this study is limited to the Mumbai Metropolitan region, the framework is designed to be scalable and adaptable to other cities of India. The limitations of the current work include reliance on open datasets, absence of real-time feeds, and parameter tuning requirements. To apply the findings, we developed a location-aware Web Application that alerts users of high-risk zones and provides access to nearby emergency services.