A New Hierarchical Ranking Aggregation Method

The ranking fusion (or aggregation), which is an important branch in multiple attribute decision making, combines multiple rankings to a single one for decision making. Many traditional ranking fusion methods are implemented through heuristic ways to reduce the computational cost. They all have their own pros and cons. In this paper, a new hierarchical ranking aggregation method is proposed. All the items are first divided into multiple ranking levels (i.e., ordered items subsets) based on the information provided by different rankings to be fused. The items in high ranking levels are ranked higher than all the items in low ranking levels in the fused ranking, therefore those items in different levels never interact each other's ranking. Then, the items in the same ranking level are further divided into multiple sub-levels if possible. In the final, the items in a sub-level which cannot be further divided are further compared and ranked in particular. Simulation results show that our new hierarchical method performs well in ranking fusion for decision making.