Spatio-temporal Modelling of Crime and Violence Trends in Mombasa County, Kenya

Criminal activities are a pervasive national security threat with far reaching effects on Kenya’s social and economic well-being. At such, criminal activities in Kenya have increased in both variety, frequency and numbers every year. Such activities span a wide range including petty theft, assault, vandalism, murder, rape, fraud, organized crime, youth gangs, kidnaps, terrorism, radicalization, and other cases. Further, the move to a devolved system of governance brings new threats of crime and violence relating to investment and urbanization of rural centers, as well as new borders and resource conflicts which may manifest in crime. Analyzing geo-coded crime data provides new insights for designing, allocating, and implementation of data driven crime prevention policies and programs. The main objective of this study is to model crime incident patterns using spatial-temporal techniques across the county with a view of informing crime prevention policy and reducing violence activities by understanding crime trends. Datasets from secondary sources were used (Kenya Police Sub County Stations). This study will apply statistical models and data science approaches to identify crime-general and crime-specific hotspots in Mombasa County in Kenya between 2019–2021. We employed Kernel Density Estimation (KDE) to determine spatial crime hotspots. We also used Moran’s I statistics to assess spatial autocorrelation in the study. To model temporal variations and predict crime occurrences in the area of study we employed Poisson regression. The findings of the study revealed that drug related offences and interpersonal violence were the most rampant crimes types. The intensity of these activities varied across the sub-counties where Mvita, Kisauni and Likoni came on top as persistent crime hotspots. With these insights the study demonstrated that there is need to incorporate Geographic Information System (GIS) with statistical modelling in understanding localized crime distribution. The study therefore recommends that the root cause of the crimes should be identified more so deployment of police officers, patrols, and surveillance infrastructure in persistent hotspots (e.g., Mvita, Kisauni, Likoni) should be enhanced.