THE PROBLEMS OF TRUST ASSESSMENT IN BIG DATA ANALYSIS
Grusho A.A., Grusho N.A., Zabezhailo M.I., Timonina E.E.
Federal Research Center “Informatics and Control” of the Russian Academy of Sciences
Annotation: The paper examines the problem of evaluation of trust to the results of complex computer analysis of data. The approach of constructing empirical dependencies based on similarity of precedents in the training sample, which has already become classical, is used. The trust approximation is based on simulating training data by random sampling from an unknown distribution. This approach implements approximate causal analysis and have advantages and disadvantages.
Keywords: Information Security, Artificial Intelligence, Causal Relationships, Contradictions in Data.