Abstract
The purpose of evidence inference is to judge truth values of the environment states under uncertain observations. This has been modeled as mathematical problems, using Bayesian inference, Dempster-Shafer theory, etc. After formalizing judgment processes in evidence inference, we found that the judgment process under uncertainty can be modeled as a Bayesian game of subjective belief and objective evidence. Another, the rational judgment involves a perfect Nash equilibrium. Evidence equilibrium is the Nash equilibrium in judgment processes. It helps us to maximize the possibility to avoid bias, and minimize the requirement for evidence. This will be helpful for the dynamic analysis of uncertain data. In this paper, we provide an Expected k-Conviction (EkC) algorithm for the dynamic data analysis based on evidence equilibrium. The algorithm uses dynamic evidence election and combination to resolve the estimation of uncertainty with time constraint. Our experimental results demonstrate that the EkC algorithm has better efficiency compared with the static evidence combination approach, which will benefit realtime decision making and data fusion under uncertainty.