Paper Contents
Abstract
Crime nowadays is not only taking place in urban area but also rural regions Crime detection is a crucial task and requires legal proofs and documentation which is a time consuming task. Legal authorities these days need modern techniques for determining crime patterns ,Hence this study has generated a high power model for detecting crimes by using XG Boost algorithm which mainly requires predictive systems that include both spatial and temporal factors to help with preventive policing. Regular hotspot mapping and statistical models often fail to capture non-linear relationships and class imbalances in crime data (Kadar et al., 2019). This research presentation offers a groundwork for spatio-temporal crime analysis and hotspot prediction using the XGBoost algorithm. The dataset belong to Delhi and Mumbai. The dataset includes geospatial coordinates (latitude, longitude) and temporal and contextual features like crime type, severity, weather conditions, and traffic density. Past research shows the potential of spatio-temporal kernel density estimation (Hu et al., 2020), ensemble approaches (Kadar et al., 2019), and systematic reviews (Butt et al., 2020). This research highlights the need to merge spatial and temporal patterns for effective crime prediction. While deep learning models like CrimeSTC give specific insights about spatio-temporal modeling (Huang et al., 2020), but in real time, its hard to actually explain how those models work. XGBoost feels like a middle groundit still gives solid predictions, but at least you can see which factors matter most through its feature importance scores. This allows law enforcement to identify key drivers of crime risk. he main aim of this study is to develop a strong predictive model that incorporates spatio-temporal attributes. It will perform feature importance analysis to find critical factors and generate interpretable hotspot maps (Learning to rank, 2020).
Copyright
Copyright © 2025 SHARON ISSAC DANIEL. This is an open access article distributed under the Creative Commons Attribution License.
