Examining Car Accident Prediction Techniques and Road Traffic Congestion: A Comparative Analysis of Road Safety and Prevention of World Challenges in Low-Income and High-Income Countries

,


Introduction
Te objective of the transportation system is to support the safe and efcient movement of people and goods. A signifcant unexpected outcome of transportation systems is road accidents with injuries and loss of life and economy. According to the World Health Organization, 1.35 million fatalities occur on the world's roads annually [1]. In addition, trafc collisions contribute signifcantly to trafc congestion, which is a serious issue afecting the society as a whole [2]. Road trafc accidents are increasingly recognized as the major problem, particularly in developing countries like Ethiopia. Te global status on road safety in 2018 indicated a continued rise in worldwide road trafc deaths, reaching 1.35 million per year in 2016, making it the leading cause of death for individuals aged 5-29. Globally, pedestrian and cyclist fatalities constitute a signifcant portion, amounting to 26% of all deaths, while users of motorized two-and threewheelers account for an additional 28%. Developing countries bear a disproportionate burden of road trafc injuries. In the African subregion, pedestrian fatalities account for over half of the total, constituting 55%. Pedestrian fatalities reach alarming levels in Ethiopia, where pedestrians account for 84% of road trafc deaths, and in Cote d'Ivoire, where pedestrians constitute 75% of road trafc fatalities [3]. Notably, Africa experiences the highest proportion of pedestrian and cyclist fatalities, contributing to 44% of all deaths. In the South-East Asia and Western Pacifc regions, the majority of fatalities occur among riders of motorized two-and three-wheelers, representing 43% and 36% of the total fatalities, respectively [4]. Despite having less than 60% of the world's motor vehicles, low-and middle-income countries bear the burden of over 90% of all road trafc deaths. Indeed, countries experienced a signifcant decline in the number of road fatalities during the initial months of 2020, with reported reductions of up to 80% [5]. Tis decrease can be attributed to the widespread implementation of lockdown measures in numerous countries as a response to the COVID-19 pandemic.
Car accidents and trafc congestion are two primary challenges faced by transportation systems worldwide. While these challenges impact individuals across all income levels, low-income countries encounter unique difculties that exacerbate these problems. Furthermore, low-income countries exhibit the highest road trafc fatality rates [1,6]. According to Heydari et al. [7], road trafc injuries represent a signifcant public health issue in low-income countries, contributing signifcantly to mortality and disability. Tere exist also signifcant gender disparities in road injury patterns, with women facing distinct risks. In the event of a car crash, women have a 47% higher likelihood of sustaining serious injuries compared to men. Furthermore, they are at a fve-fold increased risk of experiencing whiplash injuries. By 2030, it is projected that approximately 70% of the global population will reside in urban areas. Tis rapid urbanization will result in a surge in demand for urban mobility, surpassing the capacity of existing systems. Hence, the Global Plan for the Decade of Action for Road Safety 2021-2030 explicitly aims to reduce road deaths and injuries by at least 50% by 2030 [4].
Te term "accident" is commonly accepted to describe an incident involving one or more vehicles in a collision, resulting in property damage, injury, or death. Te term implies a random event without an apparent cause other than it occurring by chance [8]. Alen and Janice [9] recommended replacing "motor vehicle accident" with "motor vehicle crash" in the clinical and research language used by traumatologists. Tey argued that "crash" is a more inclusive term that encompasses a wider range of potential causes. Fatal crashes often result from driver intoxication, speeding, distraction, or carelessness and should not be considered as accidents. It is crucial to refrain from labeling such crashes as accidents, as doing so may impede victims' recovery by inhibiting their ability to assign blame and process the emotions associated with their trauma. According to the Highway Safety Manual (HSM), a crash is defned as "a set of events not under human control that leads to injury or property damage due to the collision of at least one motorized vehicle and may involve a collision with another motorized vehicle, a bicyclist, a pedestrian, or an object" [10], highlighting the various incidents that can be categorized as crashes. Road accidents are also referred to as road trafc crashes or collisions in diferent literatures, as mentioned by Wang [11]. Te distinction between a crash and an accident has a signifcant impact on determining ultimate responsibility for the car accident. When someone is at fault in a car accident, it is typically a preventable collision. Te National Safety Council defnes a preventable collision as "a collision in which the driver failed to do everything reasonable to avoid it" [12]. Tis means that the accident could have been avoided if the driver had taken appropriate action. Consequently, someone is accountable for the car crash and can be held liable for it.
Several studies have found that areas with higher population density, increased trafc, and greater urbanization tend to exhibit higher rates of trafc crashes [13]. Conversely, as vehicle numbers rapidly increase and cities expand, trafc incidents have a widespread and an escalating adverse efect on both trafc systems and the quality of social activities. Te management of trafc safety assumes a crucial role in intelligent transportation systems (ITSs). Trafc safety management encompasses a broad research domain wherein it is essential to analyze and predict the infuence of trafc incidents [14].
Trafc congestion constitutes a major problem in modern urban trafc networks if not well managed. When an excessively large number of cars attempt to utilize a single, constrained transit system, trafc congestion ensues [15]. Its devastating efects can occasionally paralyze a trafc network, consuming signifcant portions of commuters' productive hours and impeding essential services provided by incidence-intervention vehicles such as emergency and fre-fghting vehicles. Regardless of the cause, trafc congestion's impact is counterproductive and an indicative of an inefcient trafc network. Contributing factors include varying trafc conditions throughout the day and random congestion due to accidents [12].
With the societal development and improved living standards, the rapid increase in the number of vehicles has been accompanied by a rise in frequent trafc accidents. Tese accidents not only cause casualties and property losses but also disrupt trafc operations, leading to trafc congestion or even trafc breaks [16]. Te issue of trafc congestion is a common challenge faced by every country as its infrastructure develops. Consequently, the ability to forecast congestion becomes crucial for authorities in order to devise appropriate plans and undertake necessary actions to prevent its occurrence [17]. Te resulting trafc impacts have signifcant social (fatalities and injuries), economical, and environmental consequences for both developed and developing countries' economies. Te growing number of vehicles has transformed trafc, and trafc congestion has become a global issue, causing excessive delays and compromised safety [18]. A decrease in the number of crashes may be associated with the alleviation of trafc congestion in situations involving four or more lanes. However, it is important to acknowledge that having more lanes may lead drivers to feel more relaxed and drive at higher speeds, potentially resulting in more serious accidents [19].
Nonrecurrent congestion, primarily caused by trafc incidents, is widely recognized as a primary factor. Understanding when and where trafc accidents (TAs) occur on a road network is of paramount concern for transportation authorities and safety professionals [20]. Forty to ffty percent of all nonrecurring congestion can be attributed to trafc incidents. Undoubtedly, enhancing transportation safety can alleviate numerous health, fnancial, and qualityof-life issues faced by travelers [18]. Generally, the main factors afecting congestion on a road include incidents, trafc fow, and road conditions [21].
Trafc congestion not only profoundly impacts economic activities but also has adverse environmental efects, such as increasing noise pollution. Terefore, predicting trafc information in advance is crucial to enable efective and timely countermeasures for mitigation [22]. In addition, predicting crash proneness aids safety studies on urban roads, enabling the implementation of countermeasures and improvements, as well as assisting in the identifcation and prevention of crashes before they occur.
Trafc accidents are inherently spatial problems, and the machine learning (ML) models should consider various geospatial data sources and their interrelationships. Tis involves conducting numerous geo-processing operations, which can be computationally intensive. However, the primary goal of the transportation systems is to facilitate the efcient and safe movement of people and goods. Unfortunately, road accidents are a signifcant adverse outcome, leading to injuries, fatalities, trafc congestion, and economic losses. It is crucial to distinguish between a crash and an accident because the former encompasses a wide range of potential causes, such as speeding, distracted driving, or carelessness, and can be prevented. Efective trafc safety management is essential in preventing and predicting the impact of trafc incidents. Urban trafc networks face signifcant challenges related to trafc congestion, which results in delays, reduced safety, and environmental harm. Moreover, it can even paralyze trafc networks and impede the movement of emergency and frefghting vehicles. Predicting crash proneness facilitates safety studies on urban roads, enables the identifcation and prevention of accidents before they occur, and facilitates the implementation of safety measures as well as mitigating trafc congestion.
Terefore, the development of a predictive model becomes necessary to comprehend the road conditions and the distribution of road trafc accidents. Consequently, a predictive model is needed to assist road ofcials and transport managers in understanding the road conditions and the distribution of road trafc accidents, allowing them to predict and prevent future road trafc accidents and congestion. Ultimately, improving transportation safety can alleviate health, fnancial, and quality-of-life issues faced by travelers.
In general, there is a persistent and immediate requirement for research that addresses the evolving circumstances in numerous low-income countries (LICs) and enhances the performance of the transportation sector in achieving the United Nations' Sustainable Development Goals (SDGs) [23,24]. Te objective of this article is to ofer a comprehensive and current review of the global challenges associated with road trafc accidents and their resulting trafc congestion and road safety prevention mechanisms focusing on the comparison between low-income and highincome countries. As there is a lack of research conducted in low-income countries, we defne the scope and focus of the study particularly in Africa despite the fact that they bear the majority of the burden as compared to high-income countries. Te aim is to shed light on the signifcant discrepancies between these two income categories, evaluate the current prediction approaches, suggest the impact prevention mechanisms for car accidents and road network congestion, and identify future research focuses.
To ensure efcient screening, a systematic search of research papers was conducted using various search engines such as Scopus, Google, Google Scholar, Science Direct, and references from relevant articles. All data sources that presented a comprehensive analysis of road trafc accidents in both high-income and low-income countries, including their impact on trafc congestion and preventive mechanisms, were considered eligible for inclusion in the study.
Te succeeding sections of the paper are organized as follows: Section 2 provides an overview of the current state of knowledge in road trafc accidents, covering relevant literature, the contextual background, a comparative analysis of such accidents in low-income and high-income countries, methodologies for predicting car accidents, and a discussion on the relationship between road trafc accidents and congestion, including preventive measures. In Section 3, the key fndings of the subject, and in Section 4, we present in-depth discussion on the themes and sights derived from the literature review, analysis, and discussion in Sections 2 and 3, identifying gaps for further exploration. Section 5 summarizes our conclusions, including key fndings, implications, and recommendations based on the research, along with identifying possible opportunities for future research on accident and congestion prevention.

Road Trafc Accidents: A Comparison of Low-Income and
High-Income Countries of the Global Challenges. Detailed analyses of global accident statistics by the United Kingdom (UK) Transport Research Laboratory (TRL) and others indicate that fatality rates per licensed vehicle in developing countries are signifcantly higher compared to industrialized countries [25]. According to a study by Peden et al. [6] and Naci et al. [3], the distribution of road trafc deaths varies signifcantly among low-income, middle-income, and highincome countries. Te study reveals that 45% of road trafc fatalities in low-income countries occur among pedestrians, while the fgures in middle-income and high-income countries are 29% and 18%, respectively. Furthermore, the burden of road trafc injuries on vulnerable road users, such as pedestrians, difers signifcantly based on income levels. For instance, the study estimates that 227,835 pedestrians die each year in low-income countries, compared to 161,501 in middle-income countries and 22,500 in highincome countries. With economic growth, particularly in low-and middle-income countries, the number of vehicles on the roads has increased, making daily transportation more complex and dangerous [26]. Fatality rates (in relation to vehicle numbers) in the developing world, especially in African countries, can often be 20 to 30 times higher than those in European countries [27]. According to the World Health Organization (WHO) [1], developing countries witness signifcantly more fatalities from trafc accidents compared to industrialized countries, with much higher economic costs. Car accidents pose a major problem in both low-income and high-income countries, although the causes and impacts of car accidents can vary signifcantly between these two settings. In low-income countries, factors such as inadequate infrastructure, a lack of safety measures, older vehicles, and inexperienced drivers contribute to a higher rate of accidents [28]. Data unavailability in low-and middle-income countries impedes road safety improvement. Access to data is crucial for scientifc research on identifying factors causing high road risk and assessing the efectiveness of interventions [29]. Conversely, in high-income countries, despite the presence of better infrastructure and safety measures, distracted driving and speeding remain signifcant contributors to accidents [30].
Trafc accidents are one of the most signifcant issues in our lives. According to the road safety manual published by Permanent International Association of Road Congresses (PIARC) [31], fatality rates (deaths per 10,000 vehicles) are the lowest in developed countries, ranging from 1.1 to 5.0. Conversely, African countries, particularly Ethiopia, Tanzania, and Lesotho have the highest fatality rates, exceeding 100. A notable diference between high-income and lowincome countries is that road deaths have decreased by approximately 10% in Western Europe and North America since the mid-1980s, whereas they continue to rise in Africa, Asia/Pacifc, and Latin/Central America, and the Caribbean regions. Peden et al. [6] estimated that the annual losses from trafc accidents amount to 518 billion dollars, with low and average-income countries accounting for 65 billion dollars. Tis indicates that high-income countries spend 2% of their gross national product (GNP) on trafc accidents, while lowand average-income countries allocate 1 to 5.1% of their GNP. In addition, de Andrade et al. [32] demonstrated that road trafc injuries (RTIs) impose a considerable fnancial burden, particularly on developing economies. Low-and middle-income countries have the highest annual road trafc fatality rates, contributing to 80% of road trafc deaths. Moreover, RTIs are estimated to cost low-and middle-income countries over 100 billion dollars per year, equivalent to 1-2% of their GNP. Te increasing number of motor cars plays a major role in the escalating deaths and injuries resulting from trafc accidents in developing nations. As stated by Heydari et al. [7], previous road safety research in low-income countries (LICs) lacks studies on analyzing injury severity levels and understanding the factors that infuence them. Such studies, common in developed countries, help decision-makers design efective measures to reduce injuries. Terefore, further research is needed in LICs to address this gap.
Te absence or insufcient use of modern trafc management techniques can result in congested and unsafe road networks for road users, with pedestrians being particularly at risk. Unfortunately, little to no efort is made to improve conditions for these vulnerable road users. Recognizing the loss of lives and signifcant fnancial costs, scientists are committed to preventing trafc accidents in developing countries [31]. Tis problem is especially critical in lowincome countries (LICs) due to several persistent shortcomings in road safety standards, vehicle safety, and maintenance, as well as in the design and implementation of policies and safe transportation infrastructure. Figure 1 illustrates the relationship between national wealth and road death rates, based on data provided by the WHO [1].
Accurate knowledge of road crashes and their causes can help provide robust motives for the investment of appropriate and efective road safety interventions, which is particularly important in contexts where resources are limited [5]. Data limitations present substantial challenges in road safety analysis, particularly in low-income countries (LICs). Tese countries commonly face inadequate crash data due to its absence or limited availability, mainly attributed to sample size constraints and a lack of risk factors within the data. As a result, these limitations are more prevalent and frequent in LICs compared to developed nations [7,23]. Te traditional source of such information has been police road trafc crash data, although the accuracy of this data is questionable due to underreporting in all countries. To address this issue, WHO has provided estimates of the number of fatalities in each country using negative binomial modeling based on reported fatalities [1]. According to WHO's estimates, high-income countries have a higher proportion of road fatalities correctly reported, with an average of 88% in high-income countries (HICs) and 77% in middle-income countries (MICs). However, this reporting accuracy is signifcantly lower in lower-middleincome countries (LMICs) at 52% and low-income countries (LICs) at 17% of road fatalities correctly reported.
Te mortality rate is useful for comparing road safety across countries [5]. In general, developing countries experience much higher fatality rates per licensed vehicle compared to industrialized countries. Te combination of economic growth and increased vehicle use contributes to the increased risk in transportation. Among developing countries, particularly in Africa, the fatality rates are considerably higher compared to European countries, as shown in Figure 2. Te annual economic costs associated with trafc accidents are estimated to be in the billions of dollars, with low-and middle-income countries bearing the majority of this burden. However, it is important to note that the accuracy of police road trafc crash data remains questionable, with underreporting being a prevalent issue in all countries [1]. Terefore, obtaining accurate knowledge of road crashes and their causes is crucial for implementing efective road safety interventions, especially in context where resources are limited.
In general, Table 1 compares trafc accident impacts in low-income and high-income countries. It reveals that pedestrian fatalities are higher in low-income countries (Ethiopia: 84%, Cote d'Ivoire: 75%) compared to highincome countries (18%). Road trafc fatality rates are also signifcantly higher in low-income countries, and the burden of road trafc injuries primarily afects low-and middleincome countries. Furthermore, higher-income countries demonstrate better reporting accuracy for road fatalities. In summary, Table 1 highlights the disparities in trafc accident impacts between low-income and high-income countries.

Methods of Car Accident Prediction.
If drivers and pedestrians are aware of the locations and timing of collision hotspots on the roads, they are more likely to avoid them or adopt more defensive strategies when approaching [2]. Zhu [33] conducted extensive research on trafc safety warning technologies and methods using macro-forecast and microforecast approaches. Macro-forecasting involves utilizing macrodata such as the number of trafc accidents, death tolls, and the number of motor vehicles owned to forecast trafc safety warnings. To achieve this, Zhu introduced an integrated forecast method that combines time series and regression forecast methods to provide macrotrafc safety warnings. Lu et al. [34] studied the relationships between the trafc accident and factors such as road type, vehicle type, driver state, weather, and date using statistical analysis and logistic regression analysis. Based on recently collected 400 sets of accident data from 10 major roads in Beijing, they established a prediction model for accident hotspots. Te prediction model was validated and showed an approximate prediction accuracy of 86%. In their research, Abou-Amouna et al. [35] aimed to identify and analyze the factors with a signifcant impact on road accidents in Qatar and predict the total number of road accidents in 2022. Tey discussed alternative methods and found that the most applicable ones, based on previous research studies that aligned with the existing case in Qatar, were the multiple linear regression model (MLR) and artifcial neutral network (ANN) models. After analyzing these methods and comparing their fndings, they concluded that using MLR projected 355,226 accidents in 2022, while ANN projected 216,264 accidents. Terefore, they concluded that MLR provided better results than ANN due to the latter's inability to handle data with large range variations. Oyetunji et al. [36] developed a road trafc accident predictive model using the naive Bayes' model to forecast road trafc accidents in Nigeria, aiming to prevent or reduce their occurrence. Te system demonstrated reliability with 89.83% accuracy, using selected dependent variables such as road condition, road dimension, human factors, and vehicular factors. Trafc accidents resulting in signifcant damage occur frequently. Predicting future accidents in advance can be an efective solution, providing drivers with opportunities to avoid dangers or reduce damage by responding quickly. Park et al. [37] build a predictive model using the Hadoop framework to process and analyze large trafc data, along with a sampling method to address data imbalance issues. According to the experiment, the accuracy and true positive rate were 76.35% and 40.83%, respectively, showing close similarity to results from other research. Ghadge et al. [38] employed a machine learning algorithm to predict road bumps using collected data. Tey utilized an accelerometer sensor for the collection of data and GPS for plotting the location of detected potholes in Google map. Training data were analyzed using the K-means clustering algorithm, and validation was Journal of Advanced Transportation performed using the random forest classifer. Te proposed method yielded the best possible results. In another study, Yuan et al. [39] utilized big data encompassing motor vehicle crashes in Iowa from 2006 to 2013, along with a detailed road network and various weather attributes at 1-hour interval. Tey employed four classifcation models, namely, support vector machine (SVM), decision tree, random forest, and deep neural network (DNN). To address the issue of imbalanced classes, they implemented an informative negative sampling approach. In addition, they tackled the challenge of spatial heterogeneity challenge by incorporating SpatialGraph features through Eigen analysis of the road network. Te results demonstrated that employing informative sampling and integrating SpatialGraph features signifcantly enhanced the performance of all models with random forest and DNN generally outperforming the other models. Te use of machine learning algorithms for predicting car accidents has shown promising outcomes in both low-income and highincome countries. For instance, a study conducted in India employed a machine learning model to analyze factors such as road geometry and weather conditions, enabling the prediction of accidents likelihood [40]. Similarly, a study carried out in the United States employed a similar approach  to predict car accidents using real-time trafc data for accident prediction [41]. Wang et al.'s research [42] utilized foating car trajectory data and two modeling methods: a binary logistic regression model and a support vector machine (SVM) model. Tese methods were introduced and compared for predicting crash occurrences on urban expressways. Generally, the data collected from foating cars proved efective in predicting crashes on expressways and both models exhibited good performance in predicting crashes. Notably, the SVM model outperformed the binary logistic regression model signifcantly in crash prediction. Elvik [43] highlighted the use of the empirical Bayes (EB) method and accident prediction models to estimate the expected number of accidents at specifc locations, aiding in the identifcation of hazardous road areas. Strategies can be implemented to enhance the identifcation of such locations, including the development of a classifcation system for roadway elements, accident prediction models, and the identifcation of the upper percentiles of the distribution of EB-estimates of safety.
Moreover, Liu et al. [14] applied the shockwave trafc model, which can be successfully integrated with GIS spatiotemporal analysis, to predict the congestion situation for incidents and for diferent road hierarchies. Farhan Labib et al. [44] conducted a study fgure to identify signifcant factors that have a clear efect on road accidents and provide benefcial suggestions regarding this issue. Te analysis involved the use of four supervised learning techniques: decision tree, K-nearest neighbors (KNN), Naïve Bayes, and AdaBoost. Tese techniques were used to classify the severity of accidents into fatal, grievous, simple injury, and motor collision between these four categories. AdaBoost demonstrated the best performance in this classifcation task. Formosa et al. [45] developed a real-time trafc confict detection method using deep learning (DL) methodology, achieving an accuracy rate of 94%. In addition, Brühwiler et al. [46] highlighted the importance of adding geographical context features to enhance prediction performance and improve model performance for all machine learning (ML) models. Liu et al. [47] utilized machine learning algorithms, such as random forest and logistic regression, to build a prediction model for trafc accidents in a multiethnic plateau mountain area, achieving an accuracy rate of over 80%. In the study by Atumo et al. [19], ML demonstrated unique benefts that conventional approaches cannot replicate. By analyzing crash data, ML enables the identifcation of trends and patterns that are difcult to detect using other methods. Te random forest method, in particular, facilitated the ranking of predictor variables and the identifcation of sites, allowing for proactive consideration of precursor variables and surpasses traditional hot spot modeling techniques. However, in a study by Sun et al. [48], a deep learning model called long short-term memory (LSTM) was employed to extract temporal features from trafc accident data, while XGBoost was used to learn the spatial correlations between diferent features. Te proposed hybrid model, evaluated on a real-world dataset of trafc accidents in China, outperformed traditional machine learning models, such as random forest and support vector machine, as well as single models such as LSTM and XGBoost. In addition, machine learning (ML) algorithms have been utilized to develop predictive models for trafc incidents [49,50]. In Habibzadeh et al. [20], ML was identifed as an efective tool for predicting accident severity and safety solutions on rural roads. Te utilization of machine learning algorithms, such as multiple linear regression models, artifcial neural networks, random forest, and deep neural networks, has demonstrated promising results in accident prediction [45,51]. ML algorithms can analyze large amounts of trafc data, including trafc fow, weather conditions, and road geometry, to accurately predict the likelihood of accidents or trafc congestion.
Hence, various studies have suggested that predicting road accidents is a promising approach to mitigating their incidence. Tese studies have implemented data analysis and machine learning techniques to develop predictive models that analyze crucial factors such as road type, driver state, vehicle type, weather, and date. Furthermore, these studies have demonstrated that integrating spatial and temporal data, as well as applying informative sampling, can significantly enhance the models' performance. Moreover, the implementation of warning technologies that identify collision hotspots can aid drivers and pedestrians in avoiding accidents. In addition, the use of deep learning methodology has shown high accuracy in real-time trafc confict detection. It is worth noting that these studies emphasize the importance of incorporating spatial data, such as foating car trajectory data and detailed road network data, to build precise accident prediction models. Accurately predicting accidents can help mitigate their occurrence and severity by enabling drivers to steer clear of hazardous areas or take appropriate precautions when approaching them.

Trafc Accidents and Road Network Congestion: Te
Prevention. Trafc accidents and road network congestion are closely related. Congestion on roads can lead to an increased risk of trafc accidents, while accidents can also contribute to congestion and further trafc delays. In the articles by Zhu et al. [52] and Chen et al. [53], it is noted that trafc accidents are one of the main factors causing road network congestion on urban roads and increasing travel time. Te concepts of "predict and provide" and "predict and prevent," introduced by Goodwin [54], are associated with the challenges of urban trafc management. Te idea is to predict the amount of trafc and then construct sufcient road capacity to accommodate it, aiming to prevent trafc accidents and congestion on the road network. However, merely increasing roadway capacity may not be enough, as it often generates trafc. Trafc congestion maintains equilibrium by reaching a threshold where delays discourage additional peak-period vehicle trips. Expanding congested roads attracts latent demand, resulting in generated trafc from other routes, times, and modes. Tis additional peakperiod vehicle trafc is infuenced by roadway improvements that reduce user costs. Ignoring these factors can distort planning decisions, highlighting the importance of considering them in order to efectively address trafc congestion. Terefore, alternative strategies for reducing congestion prove to be more efective and cost-efcient [40]. Identifying high-crash-density road segments, commonly referred to as hotspots, is essential for enhancing road safety and promoting a secure driving environment. By pinpointing these areas, targeted interventions can be implemented to mitigate the risk and improve overall safety on the road network [55]. Tere is substantial evidence suggesting that when trafc congestion occurs downstream, the number of crashes happening upstream tends to increase. Tis phenomenon is particularly expected on high-speed roads since encountering sudden trafc stops could result in rearend collisions [56].
Te term "trafc management" is used to describe the general process of adjusting or adapting the use of existing road systems to improve trafc operations without resorting to major new construction. Trafc management usually aims to improve trafc fows, reduce accidents, improve environments, or provide better access for people and goods [27]. Hence, the absence or insufcient use of modern trafc management techniques can result in congested and unsafe road networks for road users. A study was conducted using a before-and-after methodology, comparing the number of collisions after the installation of the system. Te results showed a signifcant reduction in the number of collisions after the system was installed, suggesting that the system is efective in reducing collision risks on local roads [57]. Auclair [58] highlighted the negative impact of road trafc congestion, rating it at 54.5% in major cities worldwide. However, road trafc congestion is expected to worsen by 61.3% compared to public transportation and air pollution in the near future. Te study by Mazloh et al. [59], examined the factors that contribute to trafc congestion, including population growth, insufcient infrastructure, trafc management challenges, and inefective transportation policies. Tey concluded that a comprehensive approach involving integrated planning, policy coordination, and the utilization of advanced technologies is necessary to optimize trafc fow and alleviate congestion in urban areas.
Te study by Jiang et al. [60] built a dynamic trafc incident management system (DTIMS) for the purposes of detecting, confrming, and resolving incidents, distributing incident information, managing on-site trafc, and recording and analyzing incidents. As a major component of DTIMS, the trafc incident information management system (TIIMS) provides real-time trafc incident information to trafc managers and participants, aiding them in making quick and accurate decisions. Tis can be done by applying and utilizing geographic information system for transportation (GIS-T), this system reduces the danger and delays caused by the incidents. Te trafc information platform serves as the foundation for TIIMS, providing trafc incident information and geographic data [61].
Urbanism has a signifcant impact on trafc safety, as summarized by Tȃrîţȃ Cîmpeanu and Burlacu [62]. Terefore, urbanism and spatial planning are tools that can achieve synergy among requirements, possibilities, and road functioning, ultimately achieving sustainable development in a given area. Trough proper planning, we can ensure efcient public transport, which has a positive impact on trafc, mobility, energy savings, and the environment.
Road trafc crashes and injuries pose major health, economic, and developmental challenges for many African countries. Despite accounting for only 4% of the world's motor vehicles, African roads witness over 10% of the total collision fatalities worldwide. Unlike in developed countries, vulnerable road users, particularly Pedestrians, account for more than 40% of the total fatalities on African roads [63]. Road trafc congestion is a global phenomenon that afects cities worldwide. As Pasquale et al. [64] and Fan et al. [2] mentioned, trafc incidents are one of the primary causes of nonrecurrent trafc congestion. Congestion occurs mainly due to capacity reduction caused by lane interruptions of one or more lanes and slowdowns to observe accidents or rescue operations. Wang [11] emphasized that accidents occur when trafc is in motion, implying that no trafc would mean no accidents. Terefore, studying trafc characteristics, such as speed, density, fow, and congestion, is crucial to understanding their impacts on accidents. Tese trafc characteristics are closely interconnected, and understanding one can provide valuable insights into the other three. Trafc congestion may be benefcial in terms of road safety. Tis is based on the premise that there would be fewer fatal accidents, and the accidents that occurred would tend to be less severe due to the low average speed when congestion is present. He et al. [65], in this study, described how trafc accidents are some likely outcomes of rapid urban development and increased motorization. Tese accidents not only result in fatalities and fnancial losses but also disrupt trafc fow, leading to congestion, and compromising trafc safety. Improving road conditions, implementing efective trafc management strategies, and enhancing driver education and training are efective measures to reduce the number of trafc accidents [14].
Te relationship between congestion and safety has been widely acknowledged, but previous research has not adequately quantifed this relationship, as investigated by Te National Academies Press [66]. Congestion can lead to stalled or slowed trafc, which signifcantly increases the risk of collisions, especially when high-speed vehicles approach unexpected trafc queues. Clearly, this presents a substantial risk of collision. Tus, treatments that reduce nonrecurrent congestion can help alleviate the frequency of these conditions. As emphasized by Akhtar and Moridpour [17], forecasting congestion also enables authorities to anticipate and proactively address it by making informed plans and implementing necessary actions to prevent its occurrence.
Connected vehicles can share active safety features such as early prediction of trafc conficts, leading to mitigated risks of trafc collisions [45]. Furthermore, as evaluated by Partheeban and Hemamalini [67], adopting GIS-based transportation system management (GTSM) can identify congestion in the road network and trafc conditions can be improved [68]. Analyzing road conditions and determining the factors infuencing road trafc accidents can identify hazardous locations or black spots, serving as a decision support model for road managers and administrations to Journal of Advanced Transportation prevent or reduce future road trafc accidents and congestion on road networks. Aghajani et al. [69] proposed using GIS-based spatial statistical methods to identify and model accident hotspots, aiding decision-makers in implementing appropriate measures to mitigate road accidents.
Congestion can arise not only from road trafc accidents but also due to increasing trafc volumes. Trafc congestion maintains equilibrium by reaching a threshold where delays discourage additional peak-period vehicle trips. Expanding congested roads attracts latent demand, resulting in the generated trafc from other routes, times, and modes. Tis additional peak-period vehicle trafc is infuenced by roadway improvements that reduce user costs. Ignoring these factors can distort planning decisions, highlighting the importance of considering them in order to efectively address trafc congestion [70]. Zhu et al. [52] propose an intelligent approach to predict and prevent urban trafc congestion through real-time data analysis, considering factors such as trafc fow, road network topology, and weather conditions. Tis approach aims to proactively manage trafc to prevent accidents and reduce congestion. Intelligent transportation systems (ITS) play a crucial role in managing trafc safety and reducing accidents and congestion ( [41,71]). Adaptive trafc signal control, electronic toll collection, and intelligent speed adaptation are ITS technologies that improve trafc fow and reduce accident likelihood. ITS also facilitates the sharing of real-time trafc data with drivers, enabling them to make informed decisions and avoid congested or accident-prone areas [72]. Li et al. [73] propose an adaptive virtual lane allocation algorithm that dynamically adjusts the number and position of virtual lanes based on real-time trafc information. Tis method, considering factors like trafc fow, speed, density, and vehicle spatial distribution, demonstrates superior performance in reducing congestion and improving travel time reliability. It holds promises for implementation in urban trafc management systems to improve efciency and reduce congestion. In another study by Kavoosi et al. [40], the unstructured information management architecture (UIMA) algorithm, originally proposed for solving the spatially constrained berth schedule problem (BSP), has the potential to explore its applicability to road trafc accidents and congestion problems to use it as an efective decision support tool. Te UIMA algorithm can assist in determining the most efcient routes for emergency vehicles, helping to minimize response time, ensure efective resource utilization, and potentially save lives. Advanced optimization algorithms ofer potential solutions for accident prediction and roadway congestion problems by leveraging their capabilities in optimizing complex systems and decisionmaking processes. Tese algorithms can enhance prediction accuracy and assist in developing proactive measures to prevent accidents and alleviate congestion [51]. For instance, the UIMA algorithm can be utilized to optimize trafc management strategies and alleviate road congestion. Ambulance routing algorithms, such as NSGA-II (nondominated sorting genetic algorithm II) and MOPSO (multiobjective particle swarm optimization), can also be adapted and applied to address road trafc accidents and congestion problems prevention in ways like emergency response planning, dynamic trafc management, resource optimization and trafc fow control [74].
In summary, as indicated in Table 2 , trafc accidents and road network congestion are closely linked. Managing one can alleviate the impact of the other. Te real-time data analysis and predictive models empower trafc management systems to prevent accidents and reduce congestion on urban roads. Trafc management seeks to improve trafc fows, reduce accidents, improve environments, and provide better access for people and goods. Neglecting or underutilizing modern trafc management techniques can result in congested and unsafe road networks. Urbanism signifcantly afects trafc safety, making urbanism and spatial planning essential tools for achieving sustainable development. Low-income countries, especially in Africa, face signifcant challenges regarding road trafc crashes and injuries that constitute major health, economic, and development challenges. Trafc incidents play a major role in nonrecurrent trafc congestion, and reducing congestion through design treatments or intelligent transportation system improvements positively afects safety. Planning and funding for safety-related changes not only save lives and prevent accidents but also alleviate trafc congestion [56]. GIS and transportation research have long been intertwined [47]. Adopting GIS-based transportation system management enables the identifcation of congestion and improvement of trafc conditions in the road network solving the urban transportation problem [75]. Te UIMA algorithm, initially designed for the Berth Schedule Problem, holds potential as a decision support tool for road trafc accidents and congestion. It optimizes emergency vehicle routing, reduces response time, improves resource utilization, aids in trafc management, and mitigates congestion. Adapted ambulance routing algorithms such as NSGA-II and MOPSO enhance trafc fow control through prevention measures, including emergency response planning, dynamic trafc management, and resource optimization. Table 3 summarizes the key fndings on the connections between trafc accidents, road congestion, and prevention strategies. Developing countries, especially in Africa, have signifcantly higher fatality rates per licensed vehicle compared to industrialized nations. Low-and middle-income countries bear the greatest burden of road trafc injuries, with pedestrians being the most vulnerable users. Trafc accidents incur substantial economic costs, particularly affecting the low-and middle-income countries. Understanding the causes of road crashes is crucial for implementing efective road safety interventions, especially in resource-limited settings. Machine learning algorithms and various methods have been used to predict car accidents by considering factors like road type, weather conditions, and driver behavior. Accurate accident prediction can help reduce their frequency and severity. Trafc accidents and road congestion are closely intertwined, as congestion Table  2: Te relationship between road trafc accidents, trafc congestion, and prevention options.  (7) GIS-based transportation system management: (i) Identify and improve trafc conditions using GIS and spatial statistical methods (8) Planning and funding for safety-related changes: (i) Improve road conditions, trafc management, and driver education to reduce accidents and ease trafc Table  3: Key fndings.

Key fndings
Reference Summary Developing countries have higher fatality rates per licensed vehicle compared to industrialized countries, with signifcantly higher rates in African countries Peden et al. [6]; Naci et al. [3]; World Health Organization [1]; ITF [5] Fatality rates in developing countries, especially in Africa, are much higher than in industrialized countries. Te burden of road trafc injuries is greater in low-income countries compared to high-income countries Low-income and middle-income countries bear the majority of the economic costs associated with trafc accidents, with costs estimated to be in the billions of dollars annually Peden et al. [6]; de Andrade et al. [32] Te annual economic costs of trafc accidents are signifcantly higher in low-income countries compared to high-income countries. Tese costs constitute a substantial portion of their gross national product (GNP) Accuracy of police road trafc crash data is questionable, with underreporting being prevalent in all countries World Health Organization [1] Te accuracy of police-reported road trafc crash data is low, with underreporting being a common issue even though the degree is higher in low-income countries Accurate knowledge of road crashes and their causes is crucial for implementing efective road safety interventions, especially in resource-limited contexts increases accident risks and accidents contribute to congestion. Terefore, efective trafc management algorithmic techniques are essential to improve trafc fow, reduce accidents, and establish safer and less congested road networks free for all users.

Discussion
Te objective of this review article is to examine the worldwide challenges posed by road trafc accidents and trafc congestion particularly a comparison of low-income and high-income countries, as well as to evaluate the existing methodologies utilized in the literature for predicting accidents and addressing trafc safety and congestion prevention. In addition, the study aimed to explore the preventive mechanisms employed to mitigate the impacts of these challenges. Trafc accidents are an inevitable consequence of urban motorization and occur randomly. Nevertheless, efective management and preventive measures can help reduce the number of accidents. Investments in road infrastructure are necessary to enhance road safety, particularly for vulnerable road users, by ensuring appropriate location and design [26]. In developing countries, inefcient road space utilization, weak enforcement, uncontrolled conficts, and inadequate design of trafc and pedestrian facilities are the primary contributors to trafc congestion and road safety issues [7,51,56]. Tese factors have been identifed in various studies as signifcant challenges to road safety and trafc management [28,41,45,46,61]. However, the experience of developed countries has shown that trafc management techniques are a cost-efective way to alleviate these problems. Te signifcant negative impact of trafc congestion on road safety can be attributed to the higher speed variance among vehicles within and between lanes, as well as erratic driving behavior. Lower speeds during congested periods can help reduce the overall severity of collisions since collision severity is closely linked to speed, as stated by the Institute of Transportation Engineers [76]. By reducing trafc congestion, it is possible to improve mobility and safety simultaneously [3,15,66]. Tere are policy implications that can optimize trafc fow and improve driving behavior, such as reinforcing electronic warning signs, implementing minimum speed limits, and enforcing "average speed" on specifc stretches of the roadway. A recent study by Sachs et al. [77] demonstrates that both high-income and low-income countries can enhance their road safety systems by improving institutional frameworks, databases, and the application of new technologies and approaches. According to Formosa et al. [45], deep neural network (DNN) models hold potential for use in advanced driver assistance systems (ADAS) to develop proactive safety management strategies for improving trafc safety. However, many systems currently lack the ability to display maps, limiting the scope of trafc safety analysis.
Road trafc accidents and fatalities pose a signifcant global challenge in both low-income and high-income countries [26,46,56,60]. While low-income countries experience a higher number of fatalities, high-income countries face challenges related to distracted driving and speeding. Insufcient infrastructure, safety measures, outdated vehicles, and inexperienced drivers contribute to the high fatality rates in low-income countries. Vulnerable road users, such as pedestrians, are disproportionately afected in low-income countries. In order to develop and implement efective road safety interventions, accurate knowledge of road crashes and their causes is crucial, particularly in situations with limited funds [5,7]. However, data collection and reporting often sufer from inaccuracies, especially in LMICs and LICs [34,41]. To prevent trafc accidents and fatalities, concerted eforts from policymakers, transport planners, and the public are necessary, taking into consideration the specifc challenges faced by diferent countries.
Te use of predictive models has attracted interest among researchers in the feld of car accident prevention. Tese models suggest that trafc safety warning technologies can assist drivers and pedestrians in avoiding accident-prone areas. By utilizing macro-forecast and microforecast methods, and these technologies establish prediction models for accident hotspots based on various factors such as road type, vehicle type, driver state, weather, and date. Machine learning algorithms, including multiple linear regression models, artifcial neural networks, random forests, and deep neural networks, show promise in accident prediction and congestion mitigation. By considering factors like road type, weather conditions, driver behavior, and historical data, these algorithms enhance prediction accuracy, enabling proactive measures to prevent accidents and alleviate congestion in identifed accident-prone areas, optimize resource allocation, and improve overall trafc conditions [19,45,51]. Abou-Amouna [51] argues that the multiple linear regression model yields superior results compared to artifcial neural networks due to its ability to handle a wide range of varieties. Other studies have employed models such as naive Bayes, Hadoop framework, and classifcation models, including support vector machine, decision tree, random forest, and deep neural network [50,51,53,56]. Te inclusion SpatialGraph features and informative negative sampling can enhance the performance of all models, with random forest and deep neural network performing better than other models [19,46]. Also, machine learning methods have proven efective in addressing the limitations of spatial studies, thereby enhancing their transferability [13]. Te success of predictive models in car accident prevention has been demonstrated in both low-income and high-income countries. For instance, real-time trafc data has been utilized in the United States to predict car accidents, while foating car trajectory data and modeling methods have been employed to predict crashes on urban expressways. Predictive models enable drivers to avoid dangers or minimize damage through swift responses [56].
Geographic Information Systems for Transportation (GIS-T) have gained prominence in researching and managing real-world transportation issues, including urban trafc congestion, particularly in developed countries. However, constraints such as initial prohibitive costs and insufcient expertise have limited the widespread application of GIS-T in developing countries. Intelligent transportation systems (ITS) can facilitate the sharing of real-time trafc data and provide drivers with real-time information on trafc conditions to help them avoid congested or accident-prone areas [71]. Te signifcance of geospatial information for transport modeling is substantial, yet it is often inadequately considered in many cases, despite its potential to contribute to prevent trafc accidents and reduce road network congestion [61].
A comprehensive approach involving multiple techniques and systems is necessary to prevent trafc accidents and reduce road network congestion [56]. One crucial system is trafc management techniques, which aim to improve trafc fow, reduce accidents, and provide better accessibility for people and goods. Te implementation a dynamic trafc incident management system (DTIMS) that utilizes GIS-T and trafc information platforms can deliver real-time trafc incident information to trafc managers and participants, enabling prompt and accurate decision-making and reducing the risks and delays caused by incidents [60].
Proper urbanism and spatial planning can also contribute to preventing trafc accidents and reducing road network congestion by ensuring efcient public transport and achieving synergy among requirements, possibilities, and road functions [2,16,32,34,38,46,50,56,65]. Nonrecurrent congestion can be reduced through the application of design treatments or intelligent transportation system (ITS) improvements, leading to improved safety. Active safety features in vehicles can help mitigate the risk of trafc collisions if vehicles are connected and share early predictions of trafc conficts [56]. Te UIMA algorithm, initially developed for the Berth Schedule Problem, shows promise as a decision support tool for road trafc accidents and congestion. It optimizes emergency vehicle routing, reduces response time, improves resource utilization, aids in trafc management, and mitigates congestion [40,56]. Adapted ambulance routing algorithms like NSGA-II and MOPSO can contribute to prevention through emergency response planning, dynamic trafc management, and resource optimization, enhancing trafc fow control. Finally, algorithms such as UIMA, NSGA-II and MOPS, GIS-based transportation system management (GTSM), and machine learning algorithms can be adopted to identify trafc accidents and congestion in the road network and improve trafc conditions [17,28,40,47,50,56,78]. In general, the advantages, disadvantages, debates, and discussions surrounding the use of traditional statistical methods, spatial machine learning methods, and advanced optimization algorithms in analyzing and preventing car accidents and road network congestion are tabulated in Table 4. Traditional statistical methods ofer interpretability and robustness but have limitations in assumptions and spatial considerations [69]. Spatial machine learning methods incorporate spatial relationships but face challenges of interpretability, data requirements, and overftting [19,50,53,72]. Advanced optimization algorithms optimize resource allocation but have complexities in computation, model calibration, and practical implementation [40,64,73]. Discussions revolve around trade-ofs, integration of methods, transparency, and prioritization in optimizing objectives [9,28,63,79].

Conclusions and Future Research Directions
Road accidents and trafc congestion are among the global challenges, particularly in transportation systems. Road accidents result in fatalities, injuries, and economic losses, with developing countries experiencing the highest fatality rates. Trafc congestion is another signifcant challenge, afecting urban trafc networks and causing delays, reduced safety, and environmental harm. Te main aim of this review article was to enhance road safety and address congestion prevention mechanisms in both low-income countries (LICs) and high-income countries (HICs). Te article identifed and focused on the key challenges faced by these countries, along with preventive measures, in order to defne critical areas for future research. Te ultimate goal is to improve safety and alleviate road congestion efectively, considering the anticipated rise in trafc. To improve the understanding of trafc accidents and create efective policies to prevent them, further research is necessary. Te research studies should investigate the causes and efects of trafc crashes and examine the connection between trafc congestion and accidents. Experts in transportation and ofcials should prioritize understanding when and where accidents occur on road networks to implement safety improvement measures and prevent congestion. It is important to base techniques for identifying hazardous road locations on a profound and theoretically founded understanding of accident statistics. Geographic Information Systems (GISs) can provide real-time collection, modifcation, and update of geospatial data and attribute information, making them useful tools for analyzing trafc safety issues related to geographic location. Machine learning models that consider diferent geospatial data sources and their relationships with each other are crucial in analyzing trafc accidents and predicting accident occurrences as they are inherently spatial problems. Te use of machine learning such as deep neural network (DNN) models and advanced optimization algorithms such as UIMA, NSGA-II, and MOPS algorithms can be employed to develop proactive trafc safety management strategies for improving trafc safety and congestion. In the future, researchers should investigate the application of geospatial and machine learning techniques to analyze and predict the occurrence of accidents on road networks to avoid road trafc accident and congestion.
Research fndings suggest that reducing congestion can improve both mobility and safety simultaneously. Highincome and low-income countries can improve their road safety systems through improved institutional frameworks, improved databases, road safety engineering, and application of new technologies, approaches, and actions. However, the initial prohibitive costs and inadequate expertise have limited widespread application.
Accurate knowledge of road crashes and their causes is essential for implementing efective road safety interventions, especially where funds are limited. It is crucial to defne the problem and identify the factors that contribute to it when establishing measures to improve road trafc safety. Tis would help us comprehend the behavioral, roadrelated, and vehicle-related factors that impact the number and severity of injuries in motor vehicle accidents and enable us to identify interventions based on informed decisions. Te use of spatial analysis techniques and deep learning methodologies has the potential to enhance the process of road safety planning and decision-making in general in developing countries. Tis results in high accuracy in real-time trafc confict detection, and it is essential to incorporate spatial data to build precise accident prediction models. Terefore, predictive models that analyze crucial factors such as road type, driver state, vehicle type, weather, and date can be used to mitigate their incidence.
Tis review provides valuable insights for anyone interested in the feld of road trafc safety and ofers a comprehensive overview of the current research on accident prediction and trafc congestion prevention strategies. Evaluating various factors that have been used for accident prediction and trafc congestion analysis and the application of GIS-T in trafc accident and congestion studies are important for future studies. Future research works are better to focus also on the application of cutting-edge technologies such as machine learning algorithms, geospatial techniques, and intelligent transportation systems to improve trafc safety and reduce congestion as trafc is expected to increase in large volumes. By taking a holistic approach that includes several techniques and systems, transportation networks worldwide can address the significant problem of trafc congestion and car accidents by implementing efective road safety interventions. Further research studies prioritize investigating the trade-ofs between interpretability and accuracy when using traditional statistical methods versus spatial machine learning models. Integrating diferent methodologies, such as combining statistical methods with machine learning or optimization algorithms, that ofer a more comprehensive framework for accident analysis and prevention has to be assessed too. Furthermore, engaging in discussions on transparency, explainability, data quality in spatial machine learning, and trade-ofs in advanced optimization can improve the effectiveness of congestion prevention strategies.

Data Availability
Te data are available from the corresponding author upon request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.