Classification of crime-risk areas by using the urban environment and satellite images

Abstract

The tourism industry is very important to Thailand's economy, accounting for 11% of the nation's GDP. The COVID-19 pandemic has impacted the worldwide tourism landscape, providing challenges for the industry’s recovery. This study focuses on enhancing public safety across Bangkok, particularly within the tourism sector, which is a leading global destination and an important hub for Thailand’s tourism. Despite Bangkok’s popularity, safety remains a sensitive issue, with crime impacting both tourists and residents. This study aims to classify crime risk areas in Bangkok by integrating diverse datasets, including satellite imagery, government records, and open-source data. These findings provide guidance for improving urban safety and developing public policy.

Remote Sensing (RS) and Geographic Information Systems (GIS) are important instruments in crime risk analysis, providing for assessing spatial and environmental factors of criminal activity. Through the integration of satellite imagery and urban datasets, including Points of Interest (POI), population density, and greenspace area, RS and GIS provide the efficiency mapping and classification of high-risk areas. These technologies help with the identification of crime hotspots and contribute to the development of predictive models that enhance crime prevention strategies and urban planning.

This research adopts a data-driven approach using a crime risk record and urban environment factors influencing the crime rate by considering covering human activity, socio-economics, and infrastructure. The factors were extracted from government agencies, satellite imagery, and open data. A crime risk record extracted from the Metropolitan Police Bureau (MPB), Bangkok Metropolitan Administration (BMA), and civil complaints with label high, medium, and low risk, respectively, was assessed by five experts and defined as train data. Urban environment factors are extracted into four categories consisting of:

1). Points of Interest (POI) provide the functional characteristics of urban areas as a hub of economic and social activities.

2). Greenspace refers to tree cover, parks, and ground cover. Urban areas that are well-designed and well-maintained would reduce property and violent crime rates, but poorly maintained green spaces could attract criminal activities.

3). Demographics from the LandScan project reflect the real situation of population density, which refers to the area influx of people would increase opportunities for property crimes.

4). Nighttime lighting from the Suomi National Polar-Orbiting Partnership (SNPP) refers to the effects of increased lighting on crime rates, which are more pronounced in urban areas with higher nighttime business activity.

The four categories could be broken down into 14 factors as input data represents in raster format preparing for processing. The three machine learning methods (Extra Trees, Random Forest, and Light Gradient Boosting Machine) were applied to classify and predict crime risk areas across Bangkok. Train data, divide it into training (80%) and testing (20%), and use K-Fold Cross Validation to evaluate the performance. Model evaluating the accuracy from Precision, Recall, F1Score, ROC-AUC, and confusion matrix.

The result shows Extra Tree (ET) model outperforms other with accuracy of 97.53%. The results also reveal Business and Industrial (P8), Administration (P5), and Catering and Recreation (P2) are the most important factors that influence model accuracy, and NDVI (P13) has the least impact on the model. While analysis into specific groups of factors by the average score shows demographics is highest, followed by nighttime lights, POI, and greenspace.

The classification of crime-risk areas reveals that most areas across Bangkok are classified as low-risk, accounting for 94.89% of the total area. However, the small percentage of high and medium crime risk, accounting for 2.59% and 2.52%, respectively, are very sensitive to investigations that are thorough. The investigation shows Urban and built-up areas are the largest proportion of land use types in the total of high and medium crime risk areas, which emphasizes the correlation between urbanization, population density, and increased crime risk. Additionally, this study also analyzes specifics into eight districts where crime against tourists occurs. These districts, which attract a high influx of visitors because of famous attractions, shopping places, and nightlife, are especially vulnerable to crimes. The study reveals Pathum Wan, Watthana, and Khlong Toei districts are the highest proportion from the total area of high- and medium-risk areas in eight districts that aligns with reporting crime incidents against tourists in eight districts in 2022.

Bangkok is one of Southeast Asia's most densely populated megacities, and mass transportation plays a very important role in serving people. The Bangkok metro system, consisting of the BTS Skytrain, MRT, and Airport Rail Link, provides efficient and accessible mass transit. It facilitates mobility across commercial, residential, and tourist zone connectivity. As such, this study also analyzes specific metro stations. The classification results show that of the 150 metro stations in Bangkok within a station's 100-meter radius, 34 are in high crime risk areas and 9 are in medium crime risk areas. Furthermore, out of 25 metro stations located in districts with recorded crimes against tourists, 11 stations are classified as having a high crime risk level, while 1 station falls under the medium crime risk. These findings emphasize the relationship between metro station locations and crime risks, particularly in tourist density areas. Utilize it to develop a suitable map of the metro system for tourists and residents.

Finally, the findings indicate that areas with high POI density, population density, nighttime light, and low greenspace are more prone to crime that has an impact on safety, tourism, urbanization, social issues, transportation, and economic activity. Ensuring public safety is important for urban development, sustaining tourism, and economics. Data-driven crime assessments are helpful policymakers in enhancing public safety, promoting economic growth, and ensuring sustainable urban development.

-