It has been shown that focusing the training algorithms to the decision boundary vicinity data can improve the accuracy of several classification methods. However, previous approaches for fining decision boundary vicinity data are either computationally tedious or may perform poorly in handling problems with class overlapping. With the application of the nearest neighbor rule, this work proposes a new criterion to characterize the nearness of the training samples to the decision boundary. To demonstrate the effectiveness of the proposed approach, the proposed method is integrated with a nearest neighbor classifier design method and a neural work training approach. Experimental results show that the proposed method can reduce the size and classification error for both of the tested classifiers.

The purpose of this project is to evaluate the safety and operational differences between three left-turn treatments: permitted, protected, and protected-permitted left-turn phasing. Permitted phasing allows vehicles to turn left after yielding to any opposing vehicles; protected phasing provides an exclusive phase for vehicles to turn left that does not allow opposing vehicles; and protected-permitted phasing combines the previous phasing alternatives, allowing vehicles to turn after yielding while also providing some green time for protected left-turns.As part of evaluating the differences between these left-turn treatments, crashes before and after the change at intersections that had experienced a permanent change from one phase alternative to another were compared. The crashes that took place at these intersections were compared with the number of crashes experienced at a baseline set of intersections. A general increase in total crashes was observed for most intersections, and an increase in left-turn crashes per million entering vehicles was also observed in intersections that had experienced a change from protected to protected-permitted phasing; no other clear trends were observed.The research team also gathered simulated data using VISSIM traffic modeling software and safety data were extracted from these simulations using the Surrogate Safety Assessment Model (SSAM) created by the Federal Highway Administration to identify decision boundaries between each left-turn treatment. The simulations modeled intersections with 1-, 2-, and 3-opposing-lane configurations with permitted and protected-permitted models (split into green times of 10-, 15-, and 20-seconds) for a total of 12 different simulation models. Each model was divided into 100-225 different volume scenarios, with incremental increases in left-turn vs. opposing volumes. By exporting trajectory files from VISSIM and importing these files into SSAM, crossing conflicts for each volume combination in each model were identified and extracted. These were then entered into MATLAB to create contour maps; the contours of these maps represent the number of crossing conflicts per hour associated with different combinations of left-turn and opposing volume. Basic decision boundaries were observed in the contour maps for each model. To extract an equation to estimate each boundary, JMP Pro statistical analysis software was used to perform a linear regression analysis and develop natural log-based equations estimating the decision boundaries for each configuration and phase alternative. These equations were then charted using Excel and final decision boundaries were developed for the 1-, 2-, and 3-lane configurations between permitted and protected-permitted phasing as well as between protected-permitted and protected phasing.

No description available.