Knowledge reduction is one of the important topics in the research on rough set theory, and rough decision rules are inevitably provided with uncertainty. In this paper, a family of focal sets is constructed within the framework of evidence theory on the basis of variable precision rough set theory. Accordingly, the function of basic probability assignment is defined, and then the total information entropy is calculated for evidence theory, namely the evidence entropy. Uncertainty measure for the decision table is determined by that entropy. Based on the measure, the heuristic algorithm is proposed for decision table reduction. Finally, the experimental results show the validity of the methodology.