Rank Aggregation for Metasearch Engines Using a Self-Adapatting Genetic Algorithm with Multiple Genomic Representations
Michael L. Gargano, Maheswara Prasad Kasinadhuni · 2005
We consider the problem of combining rankings from the findings of various search engines in order to select documents based on differing and multiple criteria thus improving the results of a search. We propose using multiple genomic redundant representations in a self-adapting genetic algorithm (GA) employing various codes with different locality properties. These encoding schemes insure feasibility after performing the operations of crossover and mutation and also ensure the feasibility of the initial randomly generated population (i.e., generation 0). The GAs applied in solving this NP hard problem employ non-locality or locality representations when appropriate (i.e., the GA adapts to its current search needs) which makes the GAs more efficient [15]. Keywords:rank aggregation, metasearch, selfadapting genetic algorithm Introduction to the Problem We consider the problem of combining rankings from the findings of various search engines in order to select documents based on differing and multiple criteria thus improving the results of a search. This problem is concerned with finding a consensus ranking that “faithfully ” represents and reflects the