الفهرس 
Chapter 1 IntroductionOnly 14 pages are availabe for public view
 |  | from | 191 |  |  |
| | | from | 191 | | |
Abstract
Egypt is one of the middle-income countries belonging to the Eastern Mediterranean Region (EMR). Egypt suffers from several problems in its transportation system such as large number of crashes, traffic congestions, and pollution problems, which in turn affect the efficiency and performance of the transportation system. Although crashes are rare random events, but the majority of crashes due are predictable given the use of robust crash prediction models. Because not all highway agencies in developing countries have accurate crash data to develop their own models for predicting crashes, the focus is shifted to study the transferability of crash prediction models from developed countries. The Highway Safety Manual (HSM) developed in the United States of America (USA) is considered a principal and a comprehensive resource to road safety researchers. If the developed HSM crash prediction models can be transferred to Egypt, this will save effort, time, and money. This research examines the transferability of the crash prediction models, which is commonly referred to as Safety Performance Function (SPF), of the HSM and other ten international SPFs. Five models are from the United States of America (USA): the HSM model, and four models from the states of Virginia, North Carolina, Alabama, and Ohio. Four models are from Europe: two models from Italy, and the other two from the Netherlands, and the Czech Republic. Finally, the other two models are from Korea and Ghana. These models are transfer for the prediction of total crashes on rural multi-lane divided roads in Egypt. Four segmentation approaches are assessed in the transferability of the international SPFs, namely: (1) one-kilometer segments (S1); (2) homogenous sections (S2); (3) variable segments with respect to the presence of curvatures (S3); and (4) variable segments with respect to the presence of both curvatures and U-turns (S4). The Mean Absolute Deviation (MAD), Mean Prediction Bias (MPB), Mean Absolute Percentage Error (MAPE), Pearson χ2 statistic, and Z-score parameters are used to evaluate the performance of the transferred models. The overdispersion parameter (k) for each transferred model and each segmentation approach is recalibrated using the local data by the maximum likelihood method. Before estimating the transferability calibration factor (Cr), three methods were used to adjust the local crash prediction of the transferred models, namely: (1) the HSM default crash modification factors (CMFs); (2) local CMFs; and (3) recalibrating the constant term of the transferred model. To obtain local CMFs, the development of local SPFs for major rural arterial roads in Egypt is required. The pre-mentioned four segmentation methods were used for the SPFs development. Three modeling techniques were used for the analysis and evaluation of road safety in Egypt. The first technique is the generalized linear modeling technique (GLM). The second technique is the Artificial Neural Networks (ANN) model. The third technique is the Gaussian Process regression model (GPR).The generalized linear modeling technique was used for SPFs development using the stepwise procedure, with/without considering time effect (i.e. year-to-year variation). The Akaike Information Criterion (AIC), Pearson product-moment correlation coefficient (r12), and mean prediction bias (MBP) along with the cumulative residual (CURE) plots are used to evaluate the prediction accuracy of the proposed models. The segmentation method is found to affect the prediction accuracy of the calibrated SPFs. The calibrated SPF model using the GLM technique based on the S3 segmentation method with the time-trend model form is found superior to SPFs calibrated based on other segmentation methods. In addition, the results show that by increasing each of the pavement, and shoulder widths, the probability of crash occurrence is likely to decrease. In addition, the presences of either horizontal curves and/or accesses are most likely to reduce the probability of crash occurrence. The calibrated SPF model based on the S3 segmentation method with the time-trend model form using the Gaussian Process Regression models using the Kernel function with the same inputs as in the regression analysis yielded higher correlation coefficient value (r =0.96) compared to the ANN and the GLM models. The “recalibrated constant term of the transferred model” is found to outperform the other two transferring methods (i.e. the default HSM CMFs, and the local CMFs). Besides, the results show that the segmentation method would affect the performance of the transferability process. Moreover, the Italian SPFs based on the S1 segmentation method outperforms the HSM and all of the investigated international SPFs for transferring their models to the Egyptian rural roads.