Crimes are widespread social issues that have an impact on a nation’s reputation, economic progress, and standard of living. They cost the government substantially because law enforcement, prisons, courts, and other services are required. People are at an unsettling point where they need to reduce crime even more quickly since it is rising so rapidly, especially in Kenya.

According to the National Police Service (NPS) (2022), the total number of police-recorded offenses in Kenya has increased by more than 20%, with increases in violent offenses being considerably higher. The law enforcement stations in Nairobi, Kiambu, and Meru had the greatest percentage of recorded crimes, 9.7%, 8.9%, and 6.5%, respectively, making Kiambu the county with the second highest crime rates (NPS, 2022). Researchers can evaluate and predict the locations of crimes, as well as take proactive actions beforehand to lower these numbers.

Real-time crime prediction and widespread surveillance have the potential to drastically lower crime rates, and saving lives is ultimately what matters most (Jawadi et al., 2021). The conventional crime analysis methods have become erroneous due to the abundance of crime-related information available from law enforcement and the intricate relationships among different types of data (Kennao & Kesharwani, 2018). Moreover, other factors, like socioeconomic activities, population, and school dropouts, affect crime occurrence, making crime analysis a complex process.

Therefore, modern technologies, concepts, models, and methods for enhancing crime analytics (exploration, visualization, and evaluation) simplify crime analysis and prediction, streamlining the crime eradication process. A lot of research work was done by various researchers in this domain. Nevertheless, there has been little study done on the effects of different land use types and unemployment on crime, with most previous studies concentrating on the impact of one land use type on crime-related offenses and unemployment separately. Studies examining the association between unemployment and land use and crime in large urban areas of industrialized countries make up small quantity of empirical research on the effects of land use and facilities and unemployment on the geographical distribution of crime.

Therefore, it was necessary to comprehend such a connection domestically and to consider the effect of environmental, social, and economic factors on crime simultaneously. The main objective of this research was to predict the occurrence of crime in Ndumberi municipality.

Methodology

Study Area

Ndumberi, located in Kiambu County, Kenya, is a rapidly developing municipality with a population of about 31,359 people (2019). The area is known for its agricultural activities, notably the Ndumberi Diaries Cooperative, a major yogurt producer in Kiambu County. Improved road access, particularly the Nairobi-Kiambu highway, has enhanced connectivity to Nairobi, driving further development and attracting both residents and non-residents.

However, rapid population growth has strained resources, leading to increased crime rates. Poorly designed residential areas and inadequate lighting have created environments conducive to criminal activities. While development projects and commercial activities have drawn many, particularly youth, in search of jobs, the temporary nature of these opportunities and the closure of large farms have left many unemployed and struggling, further contributing to the rise in crime.

Method

Network analysis was conducted to determine the spatial patterns of crime in relation to road networks. Land use land cover classification was conducted to obtain Ndumberi’s land use classes (bareland, forest, agriculture land, built up, and grassland). Supervised classification was used to obtain these classes using 10-metre sentinel-2 imagery.

Visibility analysis was done to determine areas that were visible or hidden from specific vantage points, showing visible and non-visible areas that influenced criminal activities. Geocoding and data cleaning of the collected crime and population data were done manually, as the data was given with locations as texts.

Zoning (functional zoning) was then done to divide Ndumberi municipality into smaller segments, which was the foundation of this study’s crime analysis. Zonal statistics operations were used on all variables to obtain summary values of the variables per zone.

Regression was done to determine which of the above factors significantly influenced crime and the extent of their influences on crime in Ndumberi. In addition, an accuracy assessment for the regression was conducted to guarantee the credibility and reliability of the regression results.

Significant independent variables obtained from the regression model served as inputs (as explanatory variables for the crimes) into the prediction model (Random Forest). Random Forest was chosen because it captured complex relationships between variables. Moreover, crime data contained some outliers and this model was robust to outliers as it made decisions based on multiple decision trees rather than relying on a single tree, leading to more stable and accurate predictions. 70% of the crime data was used for training, while the remaining 30% was used for validation of the resultant prediction. Predicted crimes then informed the decision to choose optimal sites for the installation of CCTVs and security lights, which help eradicate crime.

Results and Conclusion

This research developed a methodological approach to analyze and predict crime spatially, revealing that traffic and road accessibility significantly impact crime rates. The study emphasized the importance of incorporating environmental solutions, such as infrastructure planning and security lighting, into crime prevention strategies.

Key factors influencing crime in Ndumberi were identified, including population density, land use and land cover (LULC), unemployment rates, traffic, and low visibility. These factors highlight the socio-economic and environmental dimensions of criminal behavior. The study found that both commercial and residential zones, along with increased built and agricultural land use, contributed to crime incidence.

The regression method used was highly effective, with an r squared of 0.894, indicating strong correlations between these factors and crime. Significant crime was predicted along road networks, particularly near the Ndumberi highway. Insufficient surveillance due to low visibility was also identified as a major factor, suggesting that increasing visibility could enhance crime prevention.