This paper addresses the challenge of discovering spatio-temporally associated vehicles involved in crimes using Automatic Number Plate Recognition (ANPR) data, which is a crucial resource in public security work for obtaining vehicle trajectories. The significance of identifying associated vehicles in the context of group-crime prevention and control is emphasized. Practical experiences reveal that criminal groups often adopt subjective strategies to avoid suspicion, leading to unique spatio-temporal association patterns such as intentional long-distance following, which differ from traditional accompanying relationships and are difficult to detect with existing methods. Oriented to the actual needs of public security work, from the perspective of group-crime, to tackle this issue, the paper first analyzes the travel patterns of criminal group vehicles and categorizes them into three main spatio-temporal association modes: close-following mode, intentional long-distance following mode, and alternative-route mode. These modes reflect the different strategies used by criminals to avoid detection, ranging from maintaining close proximity to the peer vehicle to deliberately choosing different routes. Based on these patterns, the paper develops a data model using ANPR data. The study introduces spatio-temporal constraint parameters to better capture the association relationships between vehicles. These parameters include the monitoring point time constraint (Δt_i), point accompanying number (Num_Wx), continuous point accompanying number (Con_Num_Wx), intermittent accompanying distance (d), and accompanying duration (δt).The proposed method for discovering spatio-temporally associated vehicles leverages these parameters to identify potential criminal associations. The methodology involves preprocessing ANPR data to obtain vehicle trajectories, extracting candidate accompanying vehicle sets, calculating spatio-temporal constraint parameters for each candidate vehicle, and setting thresholds for these parameters to discover associated vehicles containing different spatio-temporal patterns. Finally, taking city B as an example, the relevant ANPR data of group-crimes vehicles are used for test and analysis, and the spatio-temporal constraint parameter thresholds are quantitatively evaluated based on the historical data of group-crime cases, based on which the spatio-temporal correlation vehicle analysis of a typical case is conducted, and when comparing this paper's method with the two methods of frequent sequence mining and calculating the concomitant probability, the effectiveness of this paper's method can reach up to 87.59% on average, which is better than the the comparison methods. The results show that the method can effectively identify vehicles engaged in long-distance following and alternative-route strategies, which are often missed by traditional methods. As a result, it is able to quickly target those involved in the crime and further narrow the scope of investigation. In conclusion, the paper presents a comprehensive method for discovering spatio-temporally associated vehicles using ANPR data, significantly enhancing the ability to detect vehicles with complex association patterns. This method not only broadens the application scope of spatio-temporal association discovery but also provides new insights and technical support for public security departments in addressing group-crimes.

The timely identification of potential criminal travel routes of key surveillance individuals is a crucial research focus for public security early warning systems. Current studies often concentrate on the travel patterns and destination preferences of criminals, but there is a lack of research from the criminals’ perspective, considering the built environment and road network structure to analyze their criminal travel routes. To address this gap, a spatiotemporal analysis approach is proposed, considering criminals’ cognition of concealment. Based on the principles of criminal psychology and rational choice theory, this paper categorizes the travel patterns of criminals under the cognition of concealment into two principles: "the Principle of Minimal Exposure Risk" and "the Principle of Minimal Travel Cost". Firstly, the urban perceptual elements within a criminal's perceptual range, including safety and risk perception elements that bring exposure risks, are calculated and introduced into the choice degree model. The "Criminal Choice Degree of Roads" is proposed to measure the optimal local roads within a criminal's perceptual range. Next, using the "Length of the Route Already Traveled" as the cost function and the "Criminal Choice Degree of Roads" within the perceptual range as the heuristic function, an improved heuristic algorithm is employed to calculate the overall optimal criminal travel route. Finally, from the perspective of crime prevention and control, experiments are conducted to analyze the distribution of urban road choice degrees and the criminal travel routes of key personnel. By comparing the routes obtained by this method with those derived from existing methods and actual criminal travel routes, it is shown that the routes calculated by this proposed method are more reasonable. They are more likely to be chosen by criminals for concealment and have relatively short travel distances, without long-term exposure in public areas. The routes, in terms of distance, travel time, and the urban perceptual elements they pass through, are closer to the actual travel behaviors of criminals, verifying the rationality of this method. The research conclusion provides decision support for public security early warning efforts, emphasizing the importance of balancing travel distance and exposure risk when monitoring key personnel, and the need to allocate resources based on the distribution of urban perceptual elements and road networks to enable timely and accurate crime prevention and interception.

Studying the journey-to-crime contributes to a better understanding of the mechanisms behind crime occurrence. However, existing research has primarily focused on pre-crime travel behavior, lacking sufficient investigation into post-crime travel, specifically criminal escape behavior. This paper focuses on the escape behavior of criminals in closed non-gated communities and proposes a research methodology that combines micro-level path simulation with macro-level spatial analysis to study criminal escape behavior. The method begins by addressing the problem of limited research data through the simulation-based reconstruction of criminal escape paths. Specifically, it involves constructing an escape path selection function based on rational choice theory, quantifying the spatial environment of the community using an assignment grid map, and iteratively simulating path data using an improved ant colony algorithm on the assignment grid map. Next, the method employs spatial syntax theory to deconstruct the spatial characteristics of the community in terms of topological depth and visual integration. Finally, by integrating the criminal escape path data with the spatial characteristics of the community, the spatial logic of criminal escape behavior is explored. The study conducted an empirical research focusing on night theft cases committed by outsiders in five typical communities in Z city, and the results show that: (1) Criminals' escape distances displayed a clear distance decay characteristic, and it also shows the decay effect at both ends; (2) The unique built environment and spatial structure of the community significantly affect the criminal behavior. Communities with too many entrances and exits tend to have shorter escape distances for criminals; (3)Criminals will avoid places with high visual integration and look for more concealed spaces to escape; (4) Criminal escape behaviors conform to the rules of space topology, that is, they will choose spaces with small topological depths to escape and follow the descending path of topological depths escape. The research findings provide a reference for the planning and construction transformation of communities. For example, communities should: emphasize the layout of entrances and exits, providing convenience for residents while minimizing the number of entry and exit points; strengthen the management of areas with a moderate distance from entrances and exits, and allocate more crime prevention and control resources in this area, such as increasing security patrols; integrate landscape design with community safety considerations, such as limiting the planting of shrubs that may obstruct pedestrians' sightlines, thus increasing the visual integration of community space.

Spatiotemporal crime prediction often employs quantitative techniques such as Geographic Information Systems (GIS), geo-statistics, and big data analysis to predict the time and risk area (or location) of crime events that are more likely to occur in the future. In the era of big data, how to dynamically optimize the deployment of limited police forces and successfully improve the effectiveness of crime prevention based on data-driven crime predictions is a research focus in the field of global predictive policing. It is also a main practical direction for law enforcement agencies worldwide to implement modern proactive policing models. Traditional crime geography and spatiotemporal crime prediction methods mainly use police precincts or grids as the basic spatial analysis unit, and the analysis results are not conducive to guiding refined patrol force planning and deployment. The graph neural network based on deep learning can be combined with the topological structure of the road network at the micro scale, enabling precise crime prediction at the street scale. However, existing approaches rarely consider the impact of road weights in model prediction. In this paper, a Road Weighted Spatiotemporal Graph Convolutional Network (RW-STGCN) is constructed for street crime prediction by introducing road network accessibility and distance attenuation factors, and the model is evaluated using street theft crime data of Chicago. The results show that: (1) Compared to the spatiotemporal graph convolutional networks without considering road weights, the hit rate of the RW-STGCN increases by more than 6.5% across various road network coverage ratios (1%, 5%, 10%, and 20%), and the increase becomes more significant as the coverage ratio decreases, with a maximum increase exceeding 50%. This indicates the effectiveness and superiority of the RW-STGCN for smaller units; (2) Model ablation experiments show that the hit rate of the RW-STGCN considering road weights increases by 13.5% compared to the model result without considering road weights, and the model considering both road weights has a more significant improvement in prediction performance than the model considering only a single factor of distance attenuation weight or road network access weight, with a maximum increase of 12.9% in hit rate. This suggests that deep learning methods combined with geographic features can effectively improve the accuracy of crime prediction. The RW-STGCN is helpful for street crime prediction and can provide auxiliary decision support for law enforcement agencies to conduct scientific patrol planning and police force deployment based on road networks. In addition, it is also useful for the study of road-related urban computing problems.

The street network channels people's routine activities, which in turn affects the distribution of crime incidents. Therefore, the street micro-environment is crucial to the fine-grained understanding and explanation of the spatial distribution of street crime. In the existing research on street micro-environment extraction from street view images, semantic segmentation technique is often used to calculate the pixel proportions of various elements, without identifying individual features and objects on street. Recently, some scholars have combined semantic segmentation and object detection technologies to extract complex street environment features, generating both pixel proportions and object counts. However, no research has compared the associations between street crime and the micro-environment features extracted by the two methods. In order to explore this issue, this study used the above two methods to extract street micro-environmental features from Baidu Street View images. The first method used semantic segmentation technique to extract pixel proportions of all features. The second method combined a semantic segmentation and an object detection technology to extract the pixel proportions or counts of individual features, e.g., sidewalk, building, wall, fence, tree, and grass were measured as pixel proportions, and people and light posts on street were measured as counts. After controlling for land use mixture, concentrated disadvantage, street density, length of street segment, and facilities that attract or generate crime, zero-inflated negative binomial regression models were constructed to assess the impacts of the street micro-environmental features on street property crime, such as street theft and pickpocketing. The above two street micro-environment measurements were added to the models separately, and their influences on street property crime were then compared. The results show that: (1) compared with the conventional semantic segmentation method, adding the street micro-environment features extracted by combining semantic segmentation and object detection techniques increased the model performance by 7%. Specifically, the number of pedestrians obtained by the object detection method can better reflect the actual size of people on the street than the pixel proportion of pedestrians obtained by the semantic segmentation method, resulting a stronger association between pedestrians and street property crime. Its regression coefficient increased from 0.09 to 0.32, and the order of the absolute value of the regression coefficient increased from the tenth place to the third place; (2) the street micro-environment features extracted from street view images can effectively explain the occurrence of street property crimes. Crime targets and activity support features can significantly affect street property crime. Specifically, there were significant positive associations between the number of pedestrians on the street and street property crime, and significant negative associations between pixel proportions of sidewalks, buildings, trees, lawns, and green-rate, and the street property crime. This research adds more evidence to the literature of crime geography and environmental criminology, especially for Crime Prevention Through Environment Design (CPTED).

Crime prediction is a prerequisite for crime prevention. Forecasting crime efficiently and accurately is of great significance for improving urban management and public safety. At present, it is difficult for most of the existing studies to obtain accurate crime prediction, because they usually utilize a single machine learning or deep learning model, ignoring the spatiotemporal dependence of the crime. In this paper, we propose a spatiotemporal prediction model GAERNN based on deep learning techniques. Firstly, the GAE model is used to capture the spatial distribution characteristics of crime cases. Secondly, the features with spatial dependencies are input into the GRU model after serialization to further extract the temporal features of crime sequences. Then, the spatial and temporal prediction results of theft crime are obtained by the fully connected layer processing. Finally, we select MLP and GCN to carry out contrastive experiments by using several indicators, such as RMSE and MSE, to verify the performance of our model. The results show that our model is significantly superior to other benchmark models in spatiotemporal prediction, and it can be used to prevent and control theft crime effectively.

As a forward-looking and proactive policing mode, predictive policing has been a major innovation of modern policing reforms across the USA and European countries since it was proposed in 2008. As it does not involve the use of personal privacy data and can be integrated with police patrolling and precise crime prevention strategies, place -based spatial -temporal crime prediction has been a hot research topic and main component of policing practices. This research presents a systematic review of the progress of spatial-temporal crime prediction across the world since 2013 when the RAND Corporation released its special report on predictive policing. It contributes to the literature with the following five aspects: (1) summarizing the new trends in the field of spatiotemporal crime prediction studies in terms of the number of papers, research topics, leading scholars, and academic journals. The studies on spatial-temporal crime prediction have received extensive attention from various countries in recent years, and the research themes have shown a diversified trend. The most productive scholars are mainly from China and the USA, with the main focus on spatial-temporal crime prediction model development; (2) describing the new dynamics and progress of six basic components involved in the spatial-temporal crime prediction research, which are the prediction target, temporal scale, spatial scale, prediction method, performance evaluation measure, and practical evaluation. The four most widely studied types of crimes are theft, robbery, burglary, and motor vehicle theft. For burglary crime, the typical temporal unit for spatial-temporal prediction is 1-month; For the other three types of crime, the typical temporal unit is 1-day. For these four types of crime, the typical spatial unit is 200-meter grid. The top three models with the best prediction performance are random forest model, spatial-temporal neural network model, and Hawkes process model; (3) introducing several main commercial softwares for spatial-temporal crime prediction and global predictive policing practices; (4) investigating the relevant ethical issues and potential challenges that are embedded in each stage of practical applications, including data & algorithm biases, lack of transparency and countability mechanism; (5) prospecting future research directions in spatial-temporal crime prediction areas. This research provides a brief and panoramic image of the field of spatial-temporal crime prediction and can act as a reference for researchers and practitioners in relevant fields including crime geography, smart policing, and Policing Geographic Information System (PGIS).

In criminology, journey-to-crime describes the phenomenon that how offenders be off their residing places and search for targets in space. In view of the disadvantages in the measurement of offenders' journey-to-crime and the influencing factors, the data of residential burglary offense in X city from 2015 to 2017 are collected and the walking, riding, and driving distances from the offender's residing location to their corresponding offending location are calculated using Baidu map API service. Then, this paper uses the categorical regression model to analyze the possible impact of environmental features within the offenders' residing area and the offending area, temporal pattern of crime, and subject characteristics of offenders on the journey-to-crime. The empirical analysis shows that: ① The frequency distribution of the three fitted traffic distances obeys the distance decay effect. Among more than 50% of the occurred crimes, the traffic distances between offenders' residing locations and their corresponding offending locations are no more than 10km, and also, the results indicate that there is no significant difference between any pair of three traffic distances, which demonstrates that three types of traffic distances are moderately equal in measuring journey-to-crime cost; ② In terms of influencing factors, the journey-to-crime distance of an offender is mainly affected by the environmental features within the offending area. Specifically, the results demonstrate that the more the targets concentrated and the areas being accessible, the more adjacent offenders are strongly attracted, while the more crime concentrated, the areas more strongly attract offenders from distant places; ③ In terms of the main individual characteristics of offenders, group offenders have longer journey-to-crime distance than single offenders, which reflects that group offenders have advantages in spatial cognition. It is mainly reflected in that for offenders who commit crimes alone, their "lone wolf" behavior restricts their spatial cognition to a low level, and the offenders who commit crimes in groups can learn and accumulate more about targets through information sharing between the members. At the same time, group cooperation can also support offenders to commit crimes in remote places from their residences and help them achieve higher crime benefits. Comparing with previous work on journey-to-crime, this paper includes the diversity of travel functions of offenders in their crime-committing processes, which fills the void of current research and enhances the understanding of journey-to-crime phenomenon. Meanwhile, the work finished in this paper could also be potentially and practically applied in offenders' detection scenarios, for example, geographic profiling.

Street crimes pose threats to human life. Previous research tackling street crimes such as public thefts and provocations were often conducted at the neighborhood or even larger levels, thus ignoring the fine characteristics of the built environment at the address level, as is advocated by the theory of Crime Prevention Through Environmental Design (CPTED). It has been well established that the micro built environment at the address level has a direct or indirect impact on the occurrence of various types of crime. However, the measurement of the micro built environment, often called environmental auditing, has been challenging. Most of the previous studies relied on survey samples to characterize the built environment. They suffered from two limitations: incomplete characterization of the built environment and sparsity in spatial coverage. Baidu street view imagery is a new data source that captures built environment at the micro level. It often captures images in four different directions at a given location on the street, and it covers virtually all the drivable streets. This study used a full convolution deep learning image segmentation algorithm to extract the micro characteristics of the built environment from Baidu street view images and applied them to study crimes at the small address level. Eight variables including tree, road, sidewalk et al. were selected to show the differences of micro built environment in the study area. Bayesian logistic regression models were used to evaluate the impact of micro-built environment factors on public thefts and provocations by controlling other relevant criminological factors. Results show that the model performance was improved after adding the micro characteristics. Additionally, we found that places with more trees hence less visibility were more prone to public theft cases, and places with more roads and sidewalks were less prone to public theft cases, when compared with provocative cases. In sum, the convolution deep learning image segmentation algorithm is effective in extracting street-view-derived variables, and these variables added new dimensions to crime analysis at the micro spatial scale. This study not only contributes to the literature in crime geography, but also bring new insight for crime prevention following the principle of CPTED.

Policing strategy of crime prevention is one of the important topics in crime geography.This paper systematically reviews the development and progress of policing strategy from three aspects: the characteristics of different types of crime prevention policing strategies, crime prevention experiments and the evaluation of crime prevention effectiveness. Our research shows that: ① Community policing, problem-oriented policing , hot spots policing and intelligent policing are the four main types of crime prevention policing strategies. Community policing takes the community as a unit and reduces crime through cooperation between the police and the local residents; Problem-oriented policing aims at identifying and solving social problems through a structured approach named SARA (Scanning, Analysis, Response, Assessment). Hot spots policing makes policing plan based on crime hot spots, and intervenes crime hot spots to reduce crime. Intelligent policing uses advanced technologies such as big data and artificial intelligence for more effective policing prevention. ② Crime prevention experiments compare the changes of crime before and after the intervention of police strategy between the experimental group and the control group. Most policing experiments are mainly focused on hot spots policing, which typically reduces crime in the experimental area, but may displace crime or diffuse the benefits of crime reduction to the neighboring areas; ③ Traditional police prevention evaluation mainly considers three aspects: detection rate, public security perception and social and economic benefit. To evaluate the effect of crime displacement or benefits diffusion, scholars put forward Weighted Displacement Quotient (WDQ), Spatio-Temporal Weighted Displacement Quotient (STWDQ), Difference in Differences (DID), Propensity Score Matching (PSM),integration of DID and PSM（PSM-DID）and DID-based quadrant method to measure the space-time benefits of policing strategies. In sum,while there exist abundant international research on policing strategy about crime prevention and evaluation of their space-time benefit such research in China still remain in its infancy. In the end, this paper further discusses the direction of future research on China's policing strategy.

As the core technology of predictive policing, Spatial-Temporal (ST) prediction of crime has developed rapidly from around 2000 to the present. We introduce the basic theory of ST prediction of crime at the beginning. We regard the ST prediction method of crime as a process combining corresponding models to predict the ST distribution of crimes in the future and deconstruct it into relationships between three objects: case, ST backcloth, and individual behavior. Then, based on the input factors of prediction models, we sum up three current main methods, including ① the prediction method based on the information of cases' ST location, ② the prediction method based on the backcloth and the information of cases' ST location, and ③ the prediction method based on individual behavior, the backcloth, and the information of cases' ST location. We further summarize the mechanisms of different methods in detail respectively. In addition, we compare and analyze each method based on their applicable scenarios and predictive capacities. Finally, with the development of big data technology, we present solutions to improve current prediction methods, that are to construct a data-fusion system, refine data granularity, and integrate new types of data. For model optimization, we need to improve the ability of integrating heterogeneous data from multiple sources and balancing the interpretability and predictive ability of models.

Street contact crime refers to violations of the law committed by offenders through directly contact with victims in the street such as pickpocketing, robbery and snatch, etc, which is one of the common crimes in China. It is assumed that street contact crime is the result of interaction among motivated offenders, potential targets and absence of capable guardians. Different types of big data are employed in previous studies as ambient population to represent the potential targets which is one of the essential elements in the routine activity theory. However, these types of big data can not be applied in a micro-scale study of street contact crime because of their limitations. This study aims to fill this gap by using a new type of big data, WeChat heat map, an internet application which shows demographic distribution and changes dynamically with high spatial-temporal resolution to study the street contact crime in XT, ZG city, based on dynamic spatio-temporal distribution of potential Targets. The spatio-temporal pattern of street contact crime in XT, ZG city and their influencing factors were revealed. Street contact crime data, Points of Interest (POI) and data of house prices in XT, ZG city were used in this study as well. The whole day is divided into three intervals (wee hours: 00:00-06:59, daytime:07:00-17:59, night:18:00-23:59) and negative binomial regression models are built for the three intervals accordingly. It is demonstrated that the spatio-temporal distribution of street contact crime in XT, ZG city aggregates obviously. Street contact crime in XT, ZG city mainly concentrate in urban village and night is the peak period while daytime is the low period. The count of street contact crime in XT, ZG city reach its maximum between 22:00 and 22:59. Factors have different impacts on street contact crime from interval to interval. During the wee hours, WeChat population,KTV and leisure Club have significant positive impact on street contact crime. In the daytime, WeChat population and gym have significant positive impact on street contact crime. At night, WeChat population, restaurants, Leisure Club, bus station and distance to the nearest security department have significant positive impact on street contact crime. Others factors such as internet café, shopping mall, house prices and length of road have no significant impact on street contact crime in the whole day. WeChat population as an ambient population represent the potential targets well in routine activity theory as it has significant positive impact on street contact crime in the whole day.

Machine learning is the mainstream method for crime hotspot prediction. As a popular machine learning algorithm, the random forest algorithm is widely used in the construction of crime hotspot prediction models because of its ability of handling sparse data, and reliable predictive capability and accuracy. A number of studies use multi-source data representing the geographical environment and built environment to train and construct crime hotspot prediction models. Some are theory-driven, while others more data-driven. Most crime prediction models are global models, by constructing a single model for the entire study area. These models do not fully consider the spatial variations of crime and the built environment, as well as the varying relationship between crimes and the built environment. This paper aims to fill in this gap, using public property crime as an example to demonstrate that crime prediction models can be improved by incorporating the aforementioned spatial variations and spatially varying relationship. Firstly, according to the distribution of historical crime events and the distribution of past crime hotspots, the research area was divided into four subareas: stable high-heat grids, high-heat grids, even-hot grids, and non-hot grids. Then, according to the social disorganization theory, routine activity theory, and crime pattern theory, the three covariates including the urban village, the road network, and POI (catering, entertainment and shopping malls as crime attractors and generators) were used as the covariates representing the surrounding built environment. The random forest prediction model also used historical crime data for training and validation. Different models were created for the whole study area and each of the four subareas. The results of 26 bi-week crime hotspot prediction experiments in 2017 were compared, showing that, after adding the three covariates representing the built environment, the prediction accuracy of the entire study area, stable high-heat grids, and high-heat grids were all improved. More importantly, the subarea models were substantially more accurate than the whole model. These findings strongly endorse that incorporating spatial differentiation of crime and the built environment plays a critical role in improving the performance of the prediction models. The majority of the hotspots coincide with commercial facilities that serve as crime generators or attractors. Thus, crime prevention and control should target urban villages and the areas where road densities are high. Further, the differences in subarea based models also suggest any crime fighting strategies should be adjusted to fit each local subarea, to achieve the greatest efficiency.