Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4978841 | Accident Analysis & Prevention | 2017 | 10 Pages |
â¢The need for sensitivity analysis of road safety composite indexes is highlighted.â¢PROMETHEE-RS method was designed to find optimal composite index for the given dataset.â¢Method is based on three criterions (correlation, rank and cluster variance).â¢Method has been tested at 27 police departments in the Republic of Serbia.â¢Evaluation of method has been done for 6 composite indexes based on DEA or TOPIS.
Concept of composite road safety index is a popular and relatively new concept among road safety experts around the world. As there is a constant need for comparison among different units (countries, municipalities, roads, etc.) there is need to choose an adequate method which will make comparison fair to all compared units. Usually comparisons using one specific indicator (parameter which describes safety or unsafety) can end up with totally different ranking of compared units which is quite complicated for decision maker to determine “real best performers”. Need for composite road safety index is becoming dominant since road safety presents a complex system where more and more indicators are constantly being developed to describe it. Among wide variety of models and developed composite indexes, a decision maker can come to even bigger dilemma than choosing one adequate risk measure. As DEA and TOPSIS are well-known mathematical models and have recently been increasingly used for risk evaluation in road safety, we used efficiencies (composite indexes) obtained by different models, based on DEA and TOPSIS, to present PROMETHEE-RS model for selection of optimal method for composite index. Method for selection of optimal composite index is based on three parameters (average correlation, average rank variation and average cluster variation) inserted into a PROMETHEE MCDM method in order to choose the optimal one. The model is tested by comparing 27 police departments in Serbia.