关键词: 交通工程, 交通运行状态, 不利天气, 高速公路交织区, 改进k-prototypes, 信息熵

Abstract: This paper proposes a revised k-prototype method to improve the accuracy of the classification of traffic operational status at freeway weaving segments under adverse weather conditions by introducing the weather factors to traditional traffic flow indicators. The study determines the impact of weather on traffic flow status by analyzing the variation of traffic flow indicators under various levels of rainfall, visibility, and wind speed. The random forest model is used to select the influential variables of traffic flow operating status for each lane of the weaving sections. The information entropy is then introduced to measure the dissimilarity of the k-prototype algorithm and a clustering effectiveness evaluation indicator is used to measure the effectiveness of the status. The results show that the operation status in the lane level of the freeway weaving segments can be divided into seven categories. Corresponding to the Level of Service (LOS) at all levels in the Highway Capacity Manual, the LOS under adverse weather conditions would decrease significantly. The LOSs for lanes 1 and 3 decrease with an average of 4 levels, and the LOSs for lanes 2 and 4 decrease with an average of 3 levels. Under moderate weather conditions, each lane experiences a decrease of LOS by 2 to 3 levels. Under the same LOS, the minimum speed in lanes 1 and 3 would experience a decrease ranging from 11.2 to 17.4 km · h^{- 1} , and the minimum speed decreases in lanes 2 and 4 range from 21.2 to 27.4 km · h^{- 1} . The findings provide a theoretical basis for refined traffic management under adverse weather conditions, which helps to improve the traffic LOS at freeway weaving segments.

Key words: traffic engineering, traffic operation status, adverse weather, freeway weaving area, revised k- prototypes method, information entropy