TY - JOUR PY - 1974// TI - The traffic conflicts technique: an accident prediction method JO - Transportation research record A1 - Paddock, Richard D. SP - 1 EP - 10 VL - 486 IS - N2 - A traffic conflicts technique was developed by General Motors as a method of measuring accident potential and is based on tabulation of evasive maneuvers as evidenced by brake-light indications and lane changes. For accident potential at intersections, 20 specific conflict classifications are defined. As a result of an FHWA- financed research program, Ohio became involved in the evaluation of the GM technique. At the time that the federal program ended, Ohio decided to pursue its own evaluation of the technique. This was prompted by a conviction that the theory behind the conflicts technique was sound and by a desire to find an accident prediction technique for use in Ohio. An accident projection technique is useful if it reflects the accident trends of the subject area. Early tests indicated that the algorithm published by FHWA could not be easily calibrated for Ohio data trends. Although Ohio data were used in generation of the FHWA method of accident prediction, it was felt that the data from the states of Virginia and Washington were of such volume and different nature as to bias the resulting algorithm. During 1972 and the first half of 1973, the Ohio data base was enlarged from 196 projects (more than 400 approaches) to 410 projects providing 922 approaches, of which 611 were usable for analysis purposes. A series of regression models was applied to this enlarged data base in an attempt to find a reliable accident prediction model. As a result of this analysis, accident prediction algorithms were developed that provide a mean accuracy of plus or minus 1.1 accidents per year and a 75th percentile accuracy of plus or minus 1.8 accidents per year. In addition, substantial insight into the workings of the conflicts technique has been obtained.

Language: en

LA - en SN - 0361-1981 UR - http://dx.doi.org/ ID - ref1 ER -