Safety Performance Functions Research Articles

Safety performance functions for Two-Lane urban arterial segments

This paper presents the calibration and development process of the safety performance function for the undivided two-lane urban and suburban arterial segments in New Jersey. Data requirements, the availability of required data, and the data processing and extraction methods are presented, along with detailed results of the calibration and development process. Negative binomial, Poisson, zero-inflated Poisson and Hurdle models were generated using the development database. The best model fit was based on likelihood ratio test, AIC and BIC statistics, Vuong test and rootograms. The test database was used to calculate the calibration factor for U2 segments. The predictions of the location-specific count models were then evaluated and compared to those of calibrated Highway Safety Manual model, using the test dataset. The validation test results showed that the negative binomial and hurdle models exhibited better performance in terms of absolute residuals and absolute Pearson residual statistics. This paper also shows the impact of crash location information on analyses results, and underlines that efforts made to manually extract the missing required data can easily be offset by the inaccuracies in crash frequency databases, and the thresholds used to identify intersection related crashes.

Multivariate Approach to Peak-Period Models for Crash Types and Severities in Variable Speed Systems on Freeways

According to current research, safety performance functions (SPFs) based on traditional annual average daily traffic are less suitable for addressing crash risk in traffic-management strategies that typically operate for short time intervals (e.g., peak period). Variable speed limit (VSL)/variable advisory speed (VAS) is a traffic-management strategy which is deployed to achieve speed harmonization typically during the peak hours of weekdays. However, peak-period SPFs for VSL/VAS-implemented freeways considering different crash types and severity levels are still unexplored. Therefore, in this study the multivariate Poisson lognormal (MVPLN) method was applied to estimate the different crash-type and severity-level models. The results showed that by addressing correlations among different crash types and severity levels, MVPLN models outperformed individual univariate Poisson lognormal crash-type and -severity models. From the developed models, several traffic- and roadway-related attributes—such as volume, average speed, standard deviation of speed, segment type (merge, diverge, and weaving), higher number of lanes, and so forth—could be identified as the potential crash-risk factors. Also, the proposed models were successful in capturing the spatial effects of segments (differences between segments downstream and upstream of the location of interest) on crash frequency. Further, the implementation of VSL/VAS strategy was found to be effective in reducing rear-end crashes in the crash-type model as well as property-damage-only (PDO) and C (non-incapacitating) crash severity levels in the crash-severity model. Policymakers and practitioners could benefit from this study in incorporating safety aspects into developing VSL/VAS algorithms.

Safety Effectiveness of Rectangular Rapid Flashing Beacons Pedestrian Enhancement

A rectangular rapid flashing beacon (RRFB) is a pedestrian crosswalk countermeasure system that cautions drivers by providing them with real-time warning about the presence of pedestrians in an upcoming crosswalk. This paper investigated the safety effectiveness of existing RRFBs installed in the state of Florida between 2013 and 2018. Additionally, traffic volume, roadway characteristics, and land-use mix were included as explanatory variables. The study evaluated 154 treated sites with RRFB installations and 158 control sites with similar characteristics. Safety performance functions (SPF) were developed, and crash modification factors (CMF) were calculated using the empirical Bayes (EB) methodology for various crash types. A CMF of 0.31 for total pedestrian crashes was estimated, showing that RRFBs have the potential to reduce 69% of total pedestrian-involved crashes. CMF for total vehicular crashes and rear-end crashes were also investigated to ascertain whether RRFBs have adverse effects on rear-end crashes. CMF for total vehicular crashes was found to be 1.14, showing that RRFBs do not have positive effect in reducing total vehicular crashes. Similarly, it was observed that CMF for the rear-end crashes was 1.11, indicating RRFBs might have adverse effect on rear-end crashes because vehicles needs to slow down or stop due to the presence of pedestrians. CMFunctions for significant variables for pedestrian, rear-end, and total crash models were also estimated in this study. An economic analysis indicated that the implementation of RRFBs will yield a benefit to cost ratio (calculated over 3 years) of approximately 85.5 with a 4% discount rate.

Beyond concept: The viability of exclusive lanes for zero emission vehicles on expressways

In response to climate change issues, various measures have been implemented to decarbonize transportation. One approach is to promote zero emission vehicles which can offer rapid and effective outcomes. Accordingly, exclusive lanes for zero emission vehicles on expressways have been pursued as a measure that can be implemented immediately with minimal detrimental impacts. To assess their economic viability, this study developed a methodological framework with a benefit-cost ratio for zero emission vehicle exclusive lanes, taking multiple traffic factors into account. Along with travel time savings, a safety performance function was proposed for traffic accident probability. A case study using diverse traffic datasets from the Republic of Korea revealed greater benefits from travel time savings than costs from congestion and accidents in all scenarios. The benefit-cost ratio ranged from 2.00 to 4.08, implying that operating these lanes can be an effective strategy for promoting zero emission vehicles, which address environmental issues.

Full data imputation for freeway time-specific safety performance functions' estimation

Time-specific Safety Performance Functions (SPFs) were proposed to achieve accurate and dynamic crash frequency predictions. Unfortunately, some states do not have or archive the needed high-resolution traffic data to develop time-specific SPFs. This study proposes a novel iterative imputation method to impute the 100% missing volume and speed data from different states with similar crash rates. First, this study calculated the crash rates for 18 states and applied the One-Way Analysis of variance (ANOVA) test to group the states with similar crash rates. Second, as an example FL and VA, which both have traffic data, were used to test the proposed iterative imputation method. The results indicated that the imputed traffic data could capture the same traffic pattern as the real-collected traffic data. Further, the Mean Absolute Error (MAE) between the imputed Ln Volume and the real-collected Ln Volume for FL is only 2.47 vehicles for each segment for three hours. The MAE between the imputed Ln AvgSpeed and the real-collected Ln AvgSpeed for FL is only 1.36 mph. The Mean Absolute Percentage Error (MAPE) between the imputed Ln Volume and the real-collected Ln Volume is 11.07%. Meanwhile, the MAPE between the imputed Ln AvgSpeed and the real-collected Ln AvgSpeed is 7.40%. Finally, this study applied the proposed iterative imputation method to develop time-specific SPFs for the state without traffic data and compared the results. The results illustrated that the time-specific SPFs developed by imputed traffic data perfectly reflected the significant variables for both morning and afternoon peak models, with a prediction accuracy of 87.1% for the morning peak model.

Comprehensive Investigation of Commercial Motor Vehicle Crashes at Signalized and Unsignalized Intersections in Kentucky

This study develops safety performance functions (SPFs) for “fatal and injury” (FI) commercial motor vehicle (CMV) crashes (here, a CMV is defined as a large truck or a bus) at both signalized and unsignalized intersections in Kentucky. Five count-response regression models—negative binomial, Conway–Maxwell–Poisson, heterogeneous Conway–Maxwell–Poisson, zero-inflated Conway–Maxwell–Poisson, and zero-inflated heterogeneous Conway–Maxwell–Poisson (ZI-HTCMP)—were developed and compared at each of the selected signalized and unsignalized intersections. FI CMV intersection-related crash data in Kentucky over a period of five years (2015 to 2019) were used. Information on road-specific characteristics, for example, shoulder width, median width and type, International Roughness Index (IRI) value, and traffic volume (including annual average daily traffic [AADT] and heavy vehicle percentage [HVP]) was collected. Additionally, information on intersection-specific characteristics, for example, number of through lanes and presence of exclusive left- and right-turn lanes on major and minor roads, was collected using Google Maps’ street view time slider back to the crash year. For both signalized and unsignalized intersections, the ZI-HTCMP model (with a varying dispersion parameter) outperformed the other comparative models. With regard to signalized intersections, a speed limit of greater than 45 mph on the major road, a speed limit greater than 45 mph on the minor road, and a median on the major road wider than 3 ft were significantly associated with increased FI CMV crashes. As far as unsignalized intersections were concerned, higher major road AADT, 4-legged configuration, a speed limit greater than 45 mph on the major road, and an IRI value of greater than 100 for the major road were significantly associated with increased FI CMV crashes. The SPF results (for each of the signalized and unsignalized intersections) were then used to identify the top 10 hazardous locations for each intersection type.

Multi-vehicle safety functions for freeway weaving segments using lane-level traffic data

This study develops Safety Performance Functions (SPFs) for freeway weaving segments. Due to the coexistence of three different movements including through, merging, and divering traffic, the probability of crashes in weaving segments is higher compared to other segment types. Further, the traffic flow in this section is the most unstable. Hence, to analyze detailed traffic conditions, this study utilized lane-level traffic data. The SPFs were developed using the Poisson Lognormal (PLN) regression model technique. The results showed that different traffic parameters were significant based on the types of crashes. For the rear-end crashes model, more general traffic conditions of the weaving segment were found to be significantly associated with the crash frequency such as the natural logarithm of average speed of through lanes. Nevertheless, for the sideswipe and angle crashes models, the traffic variables which are directly related to the weaving movements were selected as significant factors such as the off-ramp volume ratio, and standard deviation of speed of the rightmost lane. The results presented in this study can be meaningful in that they can serve as a basis for the weaving segments related safety evaluation studies. In addition, the developed models’ results can be a great source to establish operational strategies to improve traffic safety on freeway weaving segments.

Developing Economic Loss-Based Thresholds for Improving Context-Specific Crash Prediction Models

Negative binomial-based safety performance functions (SPFs) have been extensively used by United States Department of Transportation professionals for predictive crash analysis. Recently, the Florida Department of Transportation (FDOT) has developed a context classification approach and incorporated it into crash prediction models, which has the potential to significantly enhance their accuracy and reliability. The additional modeling contexts and parameters make it more challenging to diagnose and remedy modeling problems, however. Particularly for roadway segments with low annual average daily traffic (AADT), short lengths, or low counts of severe crashes, the SPF models significantly underestimate the actual number of crashes. This uncertainty in SPF predictions can lead FDOT practitioners to reach misleading conclusions, such as failing to detect sites with genuinely high crash rates. This project intends to establish thresholds for certain SPF parameters to ensure reliable crash predictions are obtained across various context classes. For this purpose, we (a) developed a functional statistical model that quantifies economic loss relative prediction errors as a function of AADT volume and (b) calculated the minimum context-specific AADT threshold for each segment length group, roadway category, context classification, and crash severity combination. Employing the developed AADT thresholds confirmed up to 89% reduction in SPF prediction errors for the most represented context class. In light of the results obtained, we are able to conclude that context-specific AADT thresholds perform well in significantly reducing prediction errors for the thresholded segments and contexts on Florida roadways.

Safety Performance Functions for Traffic Signals: Phasing and Geometry

Safety Performance Functions for Traffic Signals: Phasing and Geometry

Developing Safety Performance Functions and Crash Modification Factors for Urban Roundabouts in Heterogeneous Non-Lane-Based Traffic Conditions

This study aims to develop safety performance functions (SPFs) and crash modification factors (CMFs) for the safety assessment of roundabouts as a whole and their components, especially at entry, exit, and circulating lanes. This study uses historical crash data collected over five years (2015–2019) from 21 roundabouts (78 approaches) situated in some of the major cities in India. SPF and CMF models were developed using negative binomial and cross-sectional approaches. Results show that the percentage of motorized two-wheelers, average daily traffic, central island diameter, entry angle, weaving width, departure width, number of lanes at approach, exit angle, exit width, posted speed limit, weaving length, circulating roadway width, road surface condition, number of circulatory lanes, central island height, and number of legs contribute to the crash occurrence in the roundabout vicinity. In addition to this, circulating roadway width, average daily traffic, and posted speed limit are the most influencing risk factors across the entire roundabout vicinity. Furthermore, an interesting fact is that, given the asymmetrical impact of geometric and traffic factors, the overall crash rate varies across roundabout segments. The findings of this study may be useful for updating the current design guidelines for roundabouts, considering the safety perceptive. Besides, the proposed SPF and CMF tools could be used to investigate the likelihood of crashes at roundabouts. Moreover, this model would help safety experts to implement adequate safety treatment in the vicinity of roundabouts under heterogenous non-lane-based traffic conditions.

Developing Safety Performance Functions for Commercial Motor Vehicle Crashes at Interchange Ramp Segments in Kentucky

Compared with roadway segments and intersections, the safety of interchange ramp segments has not been studied extensively, especially in the context of commercial motor vehicles (CMVs). The main objective of this study was to develop a safety performance function (SPF) tool for predicting CMV crashes occurring on interchange ramp segments. Four count models, including the negative binomial (NB), heterogeneous NB (HTNB), standard Conway–Maxwell–Poisson (CMP), and heterogeneous Conway–Maxwell–Poisson (HTCMP), were used and compared while fitting CMV crash-specific SPFs along interchange ramp segments in Kentucky. The HTCMP model, which is an extension of the standard CMP model, is a more flexible approach that handles both over-dispersed and under-dispersed crash data while exhibiting varying dispersion parameters. Five-year (2015 to 2019) CMV-related crashes along Kentucky’s ramp segments were used. The model comparison results showed that the HTCMP significantly outperformed the other three models in crash prediction accuracy and goodness-of-fit statistics (e.g., the Akaike information criterion, Bayesian information criterion, and McFadden’s Pseudo R-squared). The SPF model results using the HTCMP approach indicated that on-ramps (relative to off-ramps), ramp annual average daily traffic, ramp configuration, left shoulder width, ramp gore length, absence of left roadside barrier, and presence of other merging or diverging ramps within the ramp of interest were significantly associated with CMV crash frequency on ramp segments. Potential safety countermeasures were proposed, for example, increasing ramp gore length to be at least 730 ft (since this was associated with a reduction in CMV crashes on ramp segments).

Long-Term Performance and Safety Implications of High-Friction Surfacing Treatments

Although high-friction surface treatments (HFSTs) are a proven efficient countermeasure for the reduction of roadway departure crashes, their long-term pavement performance has yet not been fully understood with large-scale field monitoring. This paper presented and analyzed 3 years of monitoring results on 36 HFST sites located in 12 states in the US ranging from 0 to 11 years in treatment age. Taking the traffic directions and number of lanes at each site into consideration, a total of 65 HFST segments were measured. Onsite data were collected by a Grip Tester for pavement friction, the PaveVision3D laser imaging system for condition evaluation and high-resolution pavement images, and the AMES high-speed profiler for surface texture and roughness. The evolutions of HFST surface characteristics and field performance were evaluated. It was found that most HFST sites continued providing superior friction values compared with the neighboring nontreated surfaces. Cracking and surface delamination were observed as the major failure types on a few sites. The delaminated area exhibited higher roughness values and lower friction numbers. Subsequently, a multivariant friction prediction model was developed to assist in understanding its influencing factors and deterioration trends. Furthermore, a safety performance function was built for crash frequency estimation on HFST sites, through which the benefit of HFST in reducing roadway departure crashes could be further quantified.

Freeway Crash Prediction Models with Variable Speed Limit/Variable Advisory Speed

Variable speed limit (VSL) and variable advisory speed (VAS) signs are efficient, cost-effective and among the state-of-the-art active traffic management (ATM) strategies. They adopt the idea of dynamically changing posted speed limits to improve highway safety performance and operation by harmonizing traffic speed. VSL/VAS system involves changing speed limits according to real-time traffic events and weather conditions. Hence, traditional average annual daily traffic (AADT) based crash prediction models may not capture the temporal effect of traffic characteristics due to the high level of aggregation. To address this issue, short-term safety performance functions (SPFs) with aggregation levels of average annual weekday hourly traffic (AAWDHT) and average annual weekday peak traffic (AAWDPT) along with AADT-based SPFs were developed using high-resolution traffic detector and VSL/VAS operational data. In this study, the Poisson log-normal model performed well at each level of aggregation and thus is recommended for developing short-term SPFs. In line with previous studies, traffic volume and standard deviation of speed were found to be positively associated with crash frequency in all the estimated models. In addition, it was found that implementation of VSL/VAS significantly reduced crash frequency by 15.97% and 26.14% for the AAWDHT and AAWDPT models, respectively. The safety improvement was captured in the short-term models in a more distinguished way than in the highly aggregated AADT-based model. The findings of this study pave the way for practitioners and policymakers to evaluate and select important parameters for VSL/VAS strategy implementation on freeways.

Safety Performance Evaluation of Flashing Yellow Arrow: Time-of-Day Versus 24-Hour Operation

The flashing yellow arrow (FYA), as the permissive left-turn indication, has been widely implemented to mitigate driving confusion caused by the permissive steady green indication. The selection of an appropriate FYA operation type, however, remains an open-ended question as few studies have evaluated the safety effectiveness of FYA under different operation types, that is, 24-hour, time-of-day (TOD), and switching from 24-hour to TOD. Moreover, since FYA has been typically used at intersections with a single left-turn lane, its safety effectiveness at intersections with dual left-turn lanes has not been well investigated. To bridge these gaps, this paper focuses on evaluating the safety effectiveness of FYA at intersections with varying numbers of left-turn lanes under different operation types and examining whether the FYA operation type switch affects safety performance. Empirical Bayes before/after studies are conducted for 24-hour, TOD, and switching from 24-hour to TOD FYA operation types using multivariate adaptive regression splines models in comparison with traditional negative binomial models. Safety performance functions are developed for different combinations of crash types and a different number of left-turn lanes. Results show that for intersections with either a single left-turn lane or dual left-turn lanes, 24-hour and TOD FYA operation types reduce crashes by 8.76% to 50%. However, intersections with dual left-turn lanes experience a 31.2% increase in total crashes when switching from 24-hour to TOD FYA operation type, while intersections with a single left-turn lane see a 60% decrease in rear-end crashes.

Multivariate Regression of Road Segments’ Accident Data in Italian Rural Networks

Increasing traffic flows on road infrastructures and the associated comfort and safety problems have led to an increased risk of accidents for road users. To take the proper corrective actions, it is fundamental to analyze the accident phenomenon in all its aspects. The purpose of the current paper was the development of an accident prediction model for rural road segments of Friuli-Venezia Giulia (FVG) Region. The model predicts the accident frequency as a function of Annual Average Daily Traffic (AADT), segment length, and both geometrical and environmental features related to the targeted road segment. The procedure is based on the Empirical Bayes (EB) method. The statistical model used to express the road segments’ safety was the multivariate regression structure of the Safety Performance Functions. Results of a CURE plots analysis verified that the model is highly reliable in predicting the accident dataset for AADT up to 12500 vehicles per day.

Time-specific hierarchical models for predicting crash frequency of reversible and high-occupancy vehicle lanes

Time-specific Safety Performance Functions (SPFs) were proposed to achieve accurate and dynamic crash frequency predictions. This study contributes to the literature by developing time-specific SPFs for freeways that include reversible lanes (RL) and freeways that include High-Occupancy Vehicle lanes (HOV) using Microwave Vehicle Detection System (MVDS) data from Virginia, Arizona and Washington States. Variables that capture the time-specific traffic turbulence were prepared and considered in the developed SPFs. Moreover, two different hierarchical models were proposed to identify factors associated with the different crash types or severity in crash frequency prediction. The results indicated that the variables representing the volume difference between reversible and general-purpose lanes (GPL) were positively associated with crash frequency. Further, the variable that indicated the design of the access point of the reversible lane was positively associated with crash frequency. The models comparison results showed that the hierarchical models outperformed the corresponding Poisson lognormal model with lower AIC and MAE values. This study also tested the proposed hierarchical models on High-Occupancy Vehicle freeway sections and reached the same conclusion on model comparison results. The significant variables representing the logarithm of volume were found to be significant and positive with crash frequency. Moreover, the difference in average speed between the HOV lanes and GPL was also found to be positive and significant with the crash frequency. In general, this study successfully identified the factors associated with the different crash types or severity in crash frequency prediction models.

Predicting safety performance on motorways during the pandemic based on geometric design: A case study in Portugal

The mobility restrictions due to the covid-19 pandemic have created an unprecedented situation in the road system. Declared lockdowns have significantly altered the traffic volumes and perhaps also driver behavior. This study conducts research on crashes occurred on motorway A29 in Portugal before and during the disrupted mobility context. Safety performance functions were adjusted with negative binomial distribution. The segmentation was based on horizontal alignments. 5 of the 27 variables initially considered were found to be significant; namely traffic volume, segment length, curvature change rate, presence of deceleration lane and year. The quality of the fitted models was verified using cumulative residuals plots. The results showed that a traditional statistical modelling approach explains accident occurrence in a similar way before and during the pandemic.

Localizing safety performance functions for two-way STOP-controlled (TWST) three-leg intersections on rural two-lane two-way (TLTW) roadways in Alabama: A geospatial modeling approach with clustering analysis

Safety Performance Functions (SPFs) can be used to predict the number of crashes for highway facilities by site characteristics, including traffic exposures and other specific site factors. The traditional approach to developing SPFs relies on factors that are observed in the data and has an unstated assumption that the relationships between safety performance and observed factors are stationary. However, there might be factors that are not captured by the data but also have significant impacts on roadway safety performance. These factors can lead to significant unobserved heterogeneity in safety performance at different sites. Failure to capture such unobserved heterogeneity in developing SPFs may result in biases and decrease the predictive accuracy. Given the interactions between highway traffic and roadway environments, the unobserved heterogeneity is likely related to the geographic space of the highway network. This study employs a spatial modeling approach, namely Geographically Weighted Negative Binomial Regression (GWNBR), to incorporate spatial heterogeneity into SPF model estimation. The GWNBR model can generate a local SPF for every site instead of a global SPF for one entire jurisdiction (e.g., a state) from the traditional approach. Local SPFs (or l-SPFs) are high-resolution and may be difficult for practitioners to use. To support the implementation of l-SPFs, this study proposes a method to aggregate l-SPFs to various geographic levels. This study first uses the 2014–2018 geo-referenced crash data from Alabama to develop l-SPFs for two-way STOP-controlled (TWST) three-leg intersections on rural two-lane two-way (TLTW) roadways in the state. The results show that l-SPFs vary substantially across Alabama. For example, the coefficients of traffic volume (AADT) on major roads range from 0.126 to 1.203 across different areas of the state. Then, an aggregation method based on K-means clustering is demonstrated to aggregate l-SPFs to various geographic levels of interest. The l-SPFs and their aggregation provide geographic flexibility in developing countermeasures and allocating funds to improve traffic safety considering local conditions.

Research on the Safety of the Left Hard Shoulder in a Multi-Lane Highway Based on Safety Performance Function

The left hard shoulder plays an important role in the event of an emergency on the inside of a multi-lane highway, but past studies have not been able to clarify the criteria for its installation or quantify the safety impact of its installation on the left side. In order to study the influence of the left hard shoulder on the safety of vehicles traveling on multi-lane highways, based on past studies that only studied the situation of four-lane highways, this paper firstly constructs a multi-lane highway simulation model under different numbers of lanes based on the VISSIM traffic simulation and uses Surrogate Safety Assessment Model (SSAM) to study the conflict characteristics of multi-lane highway vehicles under different numbers of lanes. Based on the above findings, this paper introduces the Safety Performance Function (SPF) to construct a multi-lane freeway accident prediction model, calibrates the model by adding the indexes affected by the left side hard shoulder to the basic prediction mode, and uses the historical accident data of the Badou-Shihu section of the Guangdong Northern Second Ring Highway as the basis to study the differences in accident rates of the investigated section before and after setting the left hard shoulder. The study showed that the average Time to Collision (TTC) increased by 57.2%, Maximum Deceleration (MaxD) increased by 19.2%, and Delta Speed (DeltaS) increased by 15.3% after setting hard shoulders on the left side of multi-lane freeways, and traffic conflicts on multi-lane freeways were significantly reduced, and safety was improved considerably. In addition, the rear-end conflict rate decreased by 0.17%, 0.75%, and 4.6% after setting hard shoulders on the left side of one-way three, four, and five lanes, respectively, indicating that hard shoulders on the left side are the most effective in improving the safety of one-way five-lane freeways. The accident prediction results show that within the reasonable setting range of the left hard shoulder width (0~4 m), the accident rate decreases by about 1.5% for every 0.5 m increase if only the influence of the left hard shoulder width is considered. Without considering other factors, increasing the width of the hard shoulder on the left side can reduce the number of accidents. This indicates a significant safety improvement for a one-way five-lane highway after setting the hard shoulder on the left side, and the conclusion is consistent with the simulation results. In this paper, based on past research, the research object is extended to one-way three-, four-, and five-lane highways. The findings of this paper can help the road authorities develop specifications for installing hard shoulders on the left side of multi-lane freeways and adopt strategies to improve the traffic safety level of multi-lane freeways. In addition, the models and methods used in this paper can also help build a framework for future intelligent networked vehicle avoidance systems and promote the development of intelligent networked technologies.

Traffic Conflict Prediction at Signal Cycle Level Using Bayesian Optimized Machine Learning Approaches

This study develops non-parametric models to predict traffic conflicts at signalized intersections at the signal cycle level using machine learning approaches. Three different datasets were collected, one from Surrey, Canada, and the other two from Los Angeles and Georgia, U.S.A. From the datasets, traffic conflicts measured by modified time to collision and traffic parameters such as traffic volume, shockwave area, platoon ratio, and shockwave speed were extracted. Multilayer perceptron (MLP), support vector regression (SVR), and random forest (RF) models were developed based on the Surrey dataset, and the Bayesian optimization approach was adopted to optimize the model hyperparameters. The optimized models were applied to the Los Angeles and Georgia datasets to test their transferability, and they were also compared to a traditional safety performance function (SPF) developed using negative binominal regression. The results show that all the three Bayesian optimized machine learning models have high predictive accuracy and acceptable transferability, and the MLP model is a little better than the SVR and RF models. In addition, all three models outperform the traditional SPF with regard to predictive accuracy. The model sensitivity analysis also show that the traffic volume and shockwave area have positive effects on traffic conflicts, while the platoon ratio has negative effects.