Large-Scale Rank Aggregation from Multiple Data Sources Based D3MOPSO Method

Aggregating the search result from multiple data sources is a challenging problem in the metasearch engines. However, the ordinary methods do not have enough ability to deal with these tremendous data. Aiming at this issue, based on the big data we aim to harness the efficiency and effectiveness of the aggregation result from multiple data sources and propose an aggregation framework, which consists of aggregating the multi-users requirements and preferences ranking lists and modelling discrete multi-objective evolutionary model. Based on the DPSO algorithm, we improve and optimize its encoding scheme, initialization methods, position and velocity definition, integrating updating, turbulence operator, external archive updating strategy and leaders selection, which could address the problem of low efficiency on the large scale data sources. Extensive experiments on the public datasets, real-world datasets and synthetic simulation datasets demonstrate that our method outperforms existing state-of-the-art ranking aggregation method and multi-objective evolutionary method by the efficiency, performance and convergence.