A Probabilistic-Graphical-Model Based Approach for Representing Lineages in Uncertain Data
Wei Liu · Chinese Journal of Computers · 2011
Analyzing lineage(or called provenance) of uncertain data is to trace the origin of uncertainty based on the process of data production and evolution.To represent complex correlations and their uncertainties among uncertain data objects,and then guarantee the correctness of probability computations in lineage analysis theoretically,we study the method for representing lineages of uncertain data based on Bayesian network,an important probabilistic graphical model.Starting from the lineages' Boolean formula and the uncertain data,we propose the method to transform Boolean formulas into Bayesian network equivalently,and discuss the corresponding probabilistic semantics and properties.Case studies and experimental results show that the proposal in this paper provides an effective and extensible framework for representing data correlation and evaluating uncertainties in lineage analysis.