A ampliação acelerada da demanda por transporte, mais especificamente pelo transporte rodoviário, tem provocado um aumento expressivo no número de acidentes de trânsito nesse ambiente. Consequentemente, a redução dos acidentes de trânsito tem sido um grande desafio para os pesquisadores e gestores da área rodoviária. Porém, os acidentes de trânsito são eventos complexos se considerados os diversos fatores que podem influenciá-los. Dentro desse contexto esta dissertação apresenta um estudo de modelos de previsão de acidentes, que podem ser utilizados para a avaliação do potencial de segurança em determinados locais, identificação e classificação de localidades perigosas ou com propensão a acidentes e avaliação da eficácia de medidas de melhoria da segurança. Nessa dissertação é apresentado um levantamento teórico e metodológico dos modelos de previsão de acidentes, identificando as principais variáveis adotadas bem como as técnicas utilizadas. Para cada modelo revisado foram verificadas as principais diferenças e limitações, e ainda, a análise das variáveis mais influentes presentes nesses modelos. Após, é feita uma comparação de duas abordagens distintas para estimar modelos de previsão de acidentes. A primeira consiste em estimar a ocorrência de acidentes em segmentos da via com as mudanças de características dos elementos de infraestrutura. O segundo relaciona a frequência de acidentes para um único elemento de infraestrutura da via, chamado na literatura internacional de entidade (ex: interseção, curva, tangente, etc.), com base apenas na variável relacionada ao volume de tráfego. O estudo baseado na comparação dessas duas abordagens para a previsão de acidentes revelou que a utilização do volume de tráfego como única variável independente apresenta resultados semelhantes ou até melhores que os modelos baseados em diversos elementos de infraestrutura da rodovia. / The enlargement and the accelerated development of transportation systems, more specifically the land system, have caused the number of road accidents to increase significantly. Therefore, the reduction of road accidents has been a great challenge for researchers and managers in the field of land transportation. However, considering the various factors that may influence them, road accidents are complex events. In this context, this paper presents a study of accident prediction models that can be used to assess the safety potential in certain locations, identify and rank dangerous locations or areas prone to accidents and evaluate the effectiveness of safety improvement measures. Initially, a theoretical and methodological review of accident prediction models is presented, and both the main variables adopted and the methodologies employed are identified. The main differences between all models reviewed and their limitations are presented, and the most influential variables are analyzed. In a second moment, a comparison of two different accident prediction methods is performed. The first method consists in estimating the occurrence of accidents in road sections with changes in the characteristics of infrastructure elements. The second one relates the frequency of accidents based on a single infrastructure element (intersection, curve, tangent, etc.) based on traffic volume only. The study based on the comparison of these two methods found that the use of traffic volume as the only independent variable yields similar or even better results than the models based on various road infrastructure elements.

A ampliação acelerada da demanda por transporte, mais especificamente pelo transporte rodoviário, tem provocado um aumento expressivo no número de acidentes de trânsito nesse ambiente. Consequentemente, a redução dos acidentes de trânsito tem sido um grande desafio para os pesquisadores e gestores da área rodoviária. Porém, os acidentes de trânsito são eventos complexos se considerados os diversos fatores que podem influenciá-los. Dentro desse contexto esta dissertação apresenta um estudo de modelos de previsão de acidentes, que podem ser utilizados para a avaliação do potencial de segurança em determinados locais, identificação e classificação de localidades perigosas ou com propensão a acidentes e avaliação da eficácia de medidas de melhoria da segurança. Nessa dissertação é apresentado um levantamento teórico e metodológico dos modelos de previsão de acidentes, identificando as principais variáveis adotadas bem como as técnicas utilizadas. Para cada modelo revisado foram verificadas as principais diferenças e limitações, e ainda, a análise das variáveis mais influentes presentes nesses modelos. Após, é feita uma comparação de duas abordagens distintas para estimar modelos de previsão de acidentes. A primeira consiste em estimar a ocorrência de acidentes em segmentos da via com as mudanças de características dos elementos de infraestrutura. O segundo relaciona a frequência de acidentes para um único elemento de infraestrutura da via, chamado na literatura internacional de entidade (ex: interseção, curva, tangente, etc.), com base apenas na variável relacionada ao volume de tráfego. O estudo baseado na comparação dessas duas abordagens para a previsão de acidentes revelou que a utilização do volume de tráfego como única variável independente apresenta resultados semelhantes ou até melhores que os modelos baseados em diversos elementos de infraestrutura da rodovia. / The enlargement and the accelerated development of transportation systems, more specifically the land system, have caused the number of road accidents to increase significantly. Therefore, the reduction of road accidents has been a great challenge for researchers and managers in the field of land transportation. However, considering the various factors that may influence them, road accidents are complex events. In this context, this paper presents a study of accident prediction models that can be used to assess the safety potential in certain locations, identify and rank dangerous locations or areas prone to accidents and evaluate the effectiveness of safety improvement measures. Initially, a theoretical and methodological review of accident prediction models is presented, and both the main variables adopted and the methodologies employed are identified. The main differences between all models reviewed and their limitations are presented, and the most influential variables are analyzed. In a second moment, a comparison of two different accident prediction methods is performed. The first method consists in estimating the occurrence of accidents in road sections with changes in the characteristics of infrastructure elements. The second one relates the frequency of accidents based on a single infrastructure element (intersection, curve, tangent, etc.) based on traffic volume only. The study based on the comparison of these two methods found that the use of traffic volume as the only independent variable yields similar or even better results than the models based on various road infrastructure elements.

Das Radverkehrsaufkommen in Deutschland verzeichnete in den letzten Jahren einen Zuwachs, was sich im Umkehrschluss ebenfalls im Anstieg des Unfallgeschehens mit Radfahrendenbeteiligung widerspiegelt. Um den steigenden Unfallzahlen entgegenzuwirken, empfehlen Politik und Verbände v.a. Infrastrukturmaßnahmen zu ergreifen. Davon ausgehend untersucht die vorliegende Arbeit beispielhaft für die Stadt Dresden, wie sich einzelne Infrastrukturmerkmale auf das Unfallgeschehen zwischen Rad- und motorisiertem Verkehr auswirken. Die Datengrundlage der Untersuchung stellen dabei 548 Unfälle mit Radfahrendenbeteiligung aus den Jahren 2015 bis 2019 sowie die Merkmale von 484 Knotenpunktzufahrten dar. Da die Infrastruktur das Unfallgeschehen nicht allein determiniert, werden zudem Kenngrößen des Verkehrsaufkommens einbezogen. Um das Unfallgeschehen zu untersuchen, kommen das Random Forest-Verfahren sowie die Negative Binomialregression in Form von 'Accident Prediction Models' mit vorheriger Variablenselektion anhand des LASSO-Verfahrens zum Einsatz. Die Verfahren werden jeweils auf zwei spezielle Unfalltypen für Knotenpunkte angewandt, um differenzierte Ergebnisse zu erlangen. Der erste Unfalltyp 'Abbiege-Unfall' umfasst dabei Kollisionen zwischen einem rechtsabbiegenden und einem in gleicher oder entgegengesetzter Richtung geradeausfahrenden Beteiligten, während der zweite Unfalltyp 'Einbiegen-/Kreuzen-Unfall' Kollisionen zwischen einem vorfahrtsberechtigten Verkehrsteilnehmenden und einem einbiegenden oder kreuzenden Wartepflichtigen beinhaltet. Für den Unfalltyp 'Abbiege-Unfall' zeigen die Verfahren bspw., dass eine über den Knotenpunkt komplett oder teilweise rot eingefärbte Radfahrfurt sowie eine indirekte Führung des linksabbiegenden Radverkehrs anstelle dessen Führung im Mischverkehr höhere Unfallzahlen erwarten lässt, wobei letzteres für den untersuchten Sachverhalt irrelevant erscheint und damit auf eine Schwäche bei der Variableneinbeziehung hindeutet. Im Gegensatz dazu schätzen die Verfahren für den Unfalltyp 'Einbiegen-/Kreuzen-Unfall' bspw. höhere Unfallzahlen, wenn die Anzahl der Geradeausfahrstreifen einer Zufahrt zunimmt und wenn der Knotenpunkt durch das Verkehrszeichen Z205 bzw. eine Teil-Lichtsignalanlage anstelle der Vorschrift Rechts-vor-Links geregelt wird. Zudem zeigen die Verfahren bei beiden Unfalltypen zumeist, dass die Zahl der Unfälle ab einem bestimmten Verkehrsaufkommen weniger stark ansteigt. Dieses Phänomen ist in der Wissenschaft unter dem Namen 'Safety in Numbers-Effekt' bekannt. Ein Vergleich der Modellgüten zwischen den Unfalltypen zeigt zudem, dass beide Verfahren mit ihrem Modell des Unfalltyps 'Abbiege-Unfall' bessere Vorhersagen generieren als mit ihrem Modell des Unfalltyps 'Einbiegen-/Kreuzen-Unfall'. Weiterhin unterscheiden sich die Modellgüten nach Unfalltyp nur geringfügig zwischen beiden Verfahren, weshalb davon ausgegangen werden kann, dass beide Verfahren qualitativ ähnliche Modelle des entsprechenden Unfalltyps liefern.:1 Einleitung 2 Literaturüberblick 2.1 Safety in Numbers-Effekt 2.2 Einflussfaktoren von Radverkehrsunfällen 3 Grundlagen der Unfallforschung 3.1 Unfallkategorien 3.2 Unfalltypen 4 Datengrundlage 4.1 Unfalldaten 4.2 Infrastrukturmerkmale 4.3 Überblick über verwendete Variablen 5 Methodik 5.1 Korrelationsbetrachtung 5.2 Random Forest 5.2.1 Grundlagen 5.2.2 Random Forest-Verfahren 5.2.3 Modellgütekriterien 5.2.4 Variablenbedeutsamkeit 5.3 Negative Binomialregression 5.3.1 Grundlagen 5.3.2 Accident Prediction Models 5.3.3 Variablenselektion 5.3.4 Modellgütekriterien 5.3.5 Variablenbedeutsamkeit 5.3.6 Modelldiagnostik 6 Durchführung und Ergebnisse 6.1 Korrelationsbetrachtung 6.2 Random Forest 6.2.1 Modellgütekriterien 6.2.2 Variablenbedeutsamkeit 6.3 Negative Binomialregression 6.3.1 Variablenselektion 6.3.2 Modellgütekriterien 6.3.3 Variablenbedeutsamkeit 6.3.4 Modelldiagnostik 6.4 Vergleich beider Verfahren 6.4.1 Modellgütekriterien 6.4.2 Variablenbedeutsamkeit und Handlungsempfehlungen 6.5 Vergleich mit Literaturerkenntnissen 7 Kritische Würdigung 8 Zusammenfassung und Ausblick

In this paper, we propose a Bayesian hierarchical model for predicting accident counts in future years at sites within a pool of potential road safety hotspots. The aim is to inform road safety practitioners of the location of likely future hotspots to enable a proactive, rather than reactive, approach to road safety scheme implementation. A feature of our model is the ability to rank sites according to their potential to exceed, in some future time period, a threshold accident count which may be used as a criterion for scheme implementation. Our model specification enables the classical empirical Bayes formulation – commonly used in before-and-after studies, wherein accident counts from a single before period are used to estimate counterfactual counts in the after period – to be extended to incorporate counts from multiple time periods. This allows site-specific variations in historical accident counts (e.g. locally-observed trends) to offset estimates of safety generated by a global accident prediction model (APM), which itself is used to help account for the effects of global trend and regression-to-mean (RTM). The Bayesian posterior predictive distribution is exploited to formulate predictions and to properly quantify our uncertainty in these predictions. The main contributions of our model include (i) the ability to allow accident counts from multiple time-points to inform predictions, with counts in more recent years lending more weight to predictions than counts from time-points further in the past; (ii) where appropriate, the ability to offset global estimates of trend by variations in accident counts observed locally, at a site-specific level; and (iii) the ability to account for unknown/unobserved site-specific factors which may affect accident counts. We illustrate our model with an application to accident counts at 734 potential hotspots in the German city of Halle; we also propose some simple diagnostics to validate the predictive capability of our model. We conclude that our model accurately predicts future accident counts, with point estimates from the predictive distribution matching observed counts extremely well.

Submitted by Geandra Rodrigues (geandrar@gmail.com) on 2017-12-20T18:44:42Z No. of bitstreams: 1 lierciofeitalmottajr.pdf: 12311012 bytes, checksum: fed882b800dc53c0a497d3bda1094a0e (MD5) / Rejected by Adriana Oliveira (adriana.oliveira@ufjf.edu.br), reason: Favor corrigir no título: Higway Safety Manual on 2017-12-21T10:39:55Z (GMT) / Submitted by Geandra Rodrigues (geandrar@gmail.com) on 2017-12-21T11:54:23Z No. of bitstreams: 1 lierciofeitalmottajr.pdf: 12311012 bytes, checksum: fed882b800dc53c0a497d3bda1094a0e (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2017-12-22T11:59:40Z (GMT) No. of bitstreams: 1 lierciofeitalmottajr.pdf: 12311012 bytes, checksum: fed882b800dc53c0a497d3bda1094a0e (MD5) / Made available in DSpace on 2017-12-22T11:59:40Z (GMT). No. of bitstreams: 1 lierciofeitalmottajr.pdf: 12311012 bytes, checksum: fed882b800dc53c0a497d3bda1094a0e (MD5) Previous issue date: 2017-10-27 / No presente trabalho, é realizada a aplicação do modelo de previsão de acidentes do Higway Safety Manual – HSM, publicado pela American Association of State Highway and Transportation Officials (AASTHO) no ano de 2010, à rodovia de mão dupla e pista simples, BR–116, que corta perímetros urbanos de cidades no segmento entre Caratinga e Leopoldina, Minas Gerais. Busca-se avaliar a pertinência de sua utilização como ferramenta de sistemas de gestão de segurança rodoviária. Foram coletados dados de acidentes ocorridos nos trechos de estudo referentes ao intervalo entre 2011-2015, dados sobre o volume de tráfego, além de dados geométricos da via e das suas características ambientais. Foram analisadas as aplicações do modelo em sua condição base, calibrada e refinada pelo método empírico de Bayes. Como resultado da aplicação na condição base observou-se uma discrepância entre os valores de acidentes ocorridos e previstos pelo modelo, inviabilizando a utilização direta desta aplicação. A aplicação do modelo calibrado apresentou resultados mais próximos aos reais valores de acidentes, porém ainda não sendo possível a sua utilização como modelo de previsão de acidentes. No entanto, os valores obtidos pela aplicação do modelo do HSM refinado pelo método empírico de Bayes aproximam-se de maneira satisfatória dos valores de acidentes reais ocorridos, permitindo agora a sua utilização como modelo de previsão de acidentes, tendo sua aplicação como ferramenta de sistemas de gestão de segurança rodoviária. Apesar de ainda serem necessários estudos para melhoria da calibração e da condição base, às características especificas presentes no ambiente do perímetro urbano cortado por rodovias, o modelo aplicado neste trabalho pode ser utilizado como modelo de previsão de acidentes para estes trechos. / In the present work, the Higway Safety Manual - HSM, published by the American Association of State Highway and Transportation Officials (AASTHO), is applied to the double-lane and single-lane highway BR-116, which cuts urban in the segment between Caratinga and Leopoldina, Minas Gerais. The aim of this study was to evaluate the pertinence of its use as a tool for road safety management systems. Data were collected on accidents occurring in the study sections referring to the interval between 2011-2015, data on traffic volume, as well as geometric data of the road and their environmental characteristics. The applications of the model were analyzed in their base condition, calibrated and refined by Bayes' empirical method. As a result of the application in the base condition it was observed a discrepancy between the values of accidents occurred and foreseen by the model, making it impossible to use this application directly. The application of the calibrated model presented results closer to the actual values of accidents, but it is not yet possible to use them as an accident prediction model. However, the values obtained by applying the HSM model refined by the empirical method of Bayes approach satisfactorily the values of real accidents occurred, allowing now its use as an accident prediction model, having its use as a tool of systems of road safety management. Although still necessary studies to improve the calibration and the base condition, to the specific characteristics present in the environment of the urban perimeter cut by highways, the model applied in this work can be used as a predictive model of accidents for these stretches.

No description available.

In der Arbeit wird eine Methodik einer zusammenhängenden Analyse und modellhaften Beschreibung der Verkehrssicherheit in städtischen Hauptstraßennetzen am Beispiel der Stadt Dresden entwickelt. Die dabei gewonnenen Modelle dienen der Abschätzung von Erwartungswerten von Unfallhäufigkeiten mit und ohne Personenschaden unter Berücksichtigung der Verkehrsbeteiligungsart. Die Grundlage bilden multivariate Regressionsmodelle auf Basis verallgemeinerter linearer Modelle (GLM). Die Verwendung verallgemeinerter Regressionsmodelle erlaubt eine Berücksichtigung von Verteilungen, die besser geeignet sind, den Unfallentstehungsprozess wiederzugeben, als die häufig verwendete Normalverteilung. Im konkreten Fall werden hierzu die Poisson-Verteilung sowie die negative Binomialverteilung verwendet. Um Effekte im Hauptverkehrsstraßennetz möglichst trennscharf abbilden zu können, werden vier grundsätzliche Netzelemente differenziert und das Netz entsprechend zerlegt. Unterschieden werden neben Streckenabschnitten und Hauptverkehrsknotenpunkten auch Annäherungsbereiche und Anschlussknotenpunkte. Die Kollektive der Knotenpunkte werden ferner in signalisierte und nicht-signalisierte unterteilt. Es werden zunächst Modelle unterschiedlicher Unfallkollektive getrennt für alle Kollektive der vier Netzelemente berechnet. Anschließend werden verschiedene Vorgehensweisen für eine Zusammenfassung zu Netzmodellen entwickelt. Neben der Verwendung verkehrstechnischer und infrastruktureller Größen als erklärende Variable werden in der Arbeit auch Kenngrößen zur Beschreibung der Umfeldnutzung ermittelt und im Rahmen der Regression einbezogen. Die Quantifizierung der Umfeldnutzung erfolgt mit Hilfe von Korrelations-, Kontingenz- und von Hauptkomponentenanalysen (PCA). Im Ergebnis werden Modelle präsentiert, die eine multivariate Quantifizierung erwarteter Unfallhäufigkeiten in Hauptverkehrsstraßennetzen erlauben. Die vorgestellte Methodik bildet eine mögliche Grundlage für eine differenzierte Sicherheitsbewertung verkehrsplanerischer Variantenabschätzungen. / A methodology is developed in order to predict the number of accidents within an urban main road network. The analysis was carried out by surveying the road network of Dresden. The resulting models allow the calculation of individual expectancy values for accidents with and without injury involving different traffic modes. The statistical modelling process is based on generalized linear models (GLM). These were chosen due to their ability to take into account certain non-normal distributions. In the specific case of accident counts, both the Poisson distribution and the negative binomial distribution are more suitable for reproducing the origination process than the normal distribution. Thus they were chosen as underlying distributions for the subsequent regressions. In order to differentiate overlaying influences, the main road network is separated into four basic elements: major intersections, road sections, minor intersections and approaches. Furthermore the major and minor intersections are additionally subdivided into signalised and non-signalised intersections. Separate models are calculated for different accident collectives for the various types of elements. Afterwards several methodologies for calculating aggregated network models are developed and analysed. Apart from traffic-related and infrastructural attributes, environmental parameters are derived taking into account the adjacent building structure as well as the surrounding land-use, and incorporated as explanatory variables within the regression. The environmental variables are derived from statistical analyses including correlation matrices, contingency tables and principal components analyses (PCA). As a result, a set of models is introduced which allows a multivariate calculation of expected accident counts for urban main road networks. The methodology developed can serve as a basis for a differentiated safety assessment of varying scenarios within a traffic planning process.

In der Arbeit wird eine Methodik einer zusammenhängenden Analyse und modellhaften Beschreibung der Verkehrssicherheit in städtischen Hauptstraßennetzen am Beispiel der Stadt Dresden entwickelt. Die dabei gewonnenen Modelle dienen der Abschätzung von Erwartungswerten von Unfallhäufigkeiten mit und ohne Personenschaden unter Berücksichtigung der Verkehrsbeteiligungsart. Die Grundlage bilden multivariate Regressionsmodelle auf Basis verallgemeinerter linearer Modelle (GLM). Die Verwendung verallgemeinerter Regressionsmodelle erlaubt eine Berücksichtigung von Verteilungen, die besser geeignet sind, den Unfallentstehungsprozess wiederzugeben, als die häufig verwendete Normalverteilung. Im konkreten Fall werden hierzu die Poisson-Verteilung sowie die negative Binomialverteilung verwendet. Um Effekte im Hauptverkehrsstraßennetz möglichst trennscharf abbilden zu können, werden vier grundsätzliche Netzelemente differenziert und das Netz entsprechend zerlegt. Unterschieden werden neben Streckenabschnitten und Hauptverkehrsknotenpunkten auch Annäherungsbereiche und Anschlussknotenpunkte. Die Kollektive der Knotenpunkte werden ferner in signalisierte und nicht-signalisierte unterteilt. Es werden zunächst Modelle unterschiedlicher Unfallkollektive getrennt für alle Kollektive der vier Netzelemente berechnet. Anschließend werden verschiedene Vorgehensweisen für eine Zusammenfassung zu Netzmodellen entwickelt. Neben der Verwendung verkehrstechnischer und infrastruktureller Größen als erklärende Variable werden in der Arbeit auch Kenngrößen zur Beschreibung der Umfeldnutzung ermittelt und im Rahmen der Regression einbezogen. Die Quantifizierung der Umfeldnutzung erfolgt mit Hilfe von Korrelations-, Kontingenz- und von Hauptkomponentenanalysen (PCA). Im Ergebnis werden Modelle präsentiert, die eine multivariate Quantifizierung erwarteter Unfallhäufigkeiten in Hauptverkehrsstraßennetzen erlauben. Die vorgestellte Methodik bildet eine mögliche Grundlage für eine differenzierte Sicherheitsbewertung verkehrsplanerischer Variantenabschätzungen. / A methodology is developed in order to predict the number of accidents within an urban main road network. The analysis was carried out by surveying the road network of Dresden. The resulting models allow the calculation of individual expectancy values for accidents with and without injury involving different traffic modes. The statistical modelling process is based on generalized linear models (GLM). These were chosen due to their ability to take into account certain non-normal distributions. In the specific case of accident counts, both the Poisson distribution and the negative binomial distribution are more suitable for reproducing the origination process than the normal distribution. Thus they were chosen as underlying distributions for the subsequent regressions. In order to differentiate overlaying influences, the main road network is separated into four basic elements: major intersections, road sections, minor intersections and approaches. Furthermore the major and minor intersections are additionally subdivided into signalised and non-signalised intersections. Separate models are calculated for different accident collectives for the various types of elements. Afterwards several methodologies for calculating aggregated network models are developed and analysed. Apart from traffic-related and infrastructural attributes, environmental parameters are derived taking into account the adjacent building structure as well as the surrounding land-use, and incorporated as explanatory variables within the regression. The environmental variables are derived from statistical analyses including correlation matrices, contingency tables and principal components analyses (PCA). As a result, a set of models is introduced which allows a multivariate calculation of expected accident counts for urban main road networks. The methodology developed can serve as a basis for a differentiated safety assessment of varying scenarios within a traffic planning process.