Journal of Korean Society of Transportation ISSN:1229-1366(Print) 2234-4217(Online) 대한교통학회지

This study was intended to verify the hypothesis that there is a difference in the choice of start time by travel distance and the behavioral hypothesis that such difference in departure time results in differences in vehicle trip distribution patterns by time range using actual measurement data of navigation. In the current, when estimating traffic demand, use AAWDT (Annual Average WeekDay Traffic) estimated from household traffic survey sample data to estimate traffic pattern by applying the ratio of peak and non-peak vehicle trip. However, it is thought that estimating the concentration ratio of roads with uniform rate of peak and non-peak regardless of travel distance (travel time) to analyze the network path will differ significantly from the actual traffic patterns. This study proposes a more realistic method of estimating by reflecting the traffic patterns which difference in peak and non-peak time due to differences in departure time by distance using actual measurement data of navigation. in the network analysis. This study use SK T-map data, which has the highest share of navigation service in Korea, of 261 day on weekdays during one year. SAS (statistical package program) and RapidMiner (data mining platform) were used as analysis tools to construct k-means clustering. Based on actual navigation data, interregional travel distance was divided into seven groups to have similar characteristics, and the average traffic rate was calculated by the distance groups. By grouping and simplifying similar traffic patterns such as peak, non-peak and night using 24-hours traffic concentration rate data from derived seven distance group, a study was conducted on how to most realistically express traffic patterns. In other word, using the 24-hour traffic concentration rate of seven distance groups, through the k-means clustering analysis from k=2 to k=5, the time zone with similar traffic pattern of concentration was divided into one group. The results of BIC, Elbow, Silhouette, correlation coefficient analysis by k-means clustering result, slope, and value of non-segmented simple regression model were compared according to k value. As a result of this study, it was found that it is best to classified traffic pattern by time into three groups. In addition, it proposed a method of building OD data to enable network analysis that reflected the classification of travel distance and traffic concentration ratio patterns obtained from the research results. It is expected that a more realistic analysis of the traffic patterns of time bands on roads where short, medium and long distance traffic is mixed when performing a traffic assignment based on OD data which reflecting differences in departure time by travel distance.