Development and validation of volume-to-capacity based accident prediction models

This study is focused on the relationship between accidents and the traffic/geometric capacity of highways. By using regression analyses, this research is conducted to quantify the effects of traffic-to-capacity associations upon the number of accidents on different highway types. A concept of modified capacity is introduced where some variables such as surface rates, surface types, shoulder condition, inside shoulder type and conditions, and others are taken into consideration together with the main explanatory power, namely daily V$\sb{-}$C ratios.An analysis of section lengths is conducted to investigate the influence of section lengths on the composition and explanatory power of the prediction models. For such, several section length cutoff points are analyzed and, based on the model's variations, a minimum section length is recommended for each highway type. The data sets are then constituted of sections with lengths equal or greater than the suggested cutoff points.Prediction models are generated based on four years of data analysis (1985, 1986, 1987, and 1988), where yearly, common-section, and averaged models are created. The averaged models are validated based on other averaged three-year models which are generated from the data sets for the years 1989, 1990, and 1991. The model validation is also based on existing models when prediction models for the specific highway types hereby analyzed exist from other authors.Following, daily V$\sb{-}$C and six-year averaged models are created according to the capacity calculated on the Highway Capacity Manual and the geometric models developed for each highway type. Capacity is calculated on a daily basis to comply with the daily characteristics of the traffic flow.Geometric and Daily V$\sb{-}$C accident prediction models are then recommended for rural and urban interstate highways, rural and urban two-lane highways, and urban multilane divided and undivided highways.U of I OnlyETDs are only available to UIUC Users without author permissio