Metasearch aggregation using linear programming and neural networks

A metasearch engine aggregates the retrieved results of multiple search engines for a submitted query. The purpose of this paper is to formulate a metasearch aggregation using linear programming and neural networks by incorporating the importance weights of the involved search engines. A two-stage methodology is introduced where the importance weights of individual search engines are determined using a neural network model. The weights are then used by a linear programming model for aggregating the final ranked list. The results from the proposed method are compared with the results obtained from a simple model that assumes subjective weights for search engines. The comparison of the two sets of results shows that neural network-based linear programming model is superior in optimising the relevance of aggregated results.