Bayesian conflicting Web data credibility algorithm
Wang Ji kui · Lanzhou University Institutional Repository · 2016
The key of Web data fusion is to judge the credibility of conflicting Web data. The credibility of observing values was only rectified by their similarity relations, while the inclusion relations was ignored in existed algorithms. To solve this problem, the concept of inclusion degree was defined based on the characteristic analysis of the observing values; a modified model was proposed to rectify the credibility of observing values using inclusion degree was proposed. Taking the observing values as random variables, the reliability problem of the observation values could be attributed to a posterior probability distribution problem. Bayesian theory was adopted to define the conception of data source credibility, to derive the relationship model for data source credibility and observing value credibility, and to propose the Bayesian conflicting Web data credibility algorithm: DataCredibility. The experiment results show that the proposed DataCredibility algorithm achieves better accuracy, recall rate and F1 measure value compared with the baseline algorithms. © 2016, Zhejiang University Press. All right reserved.