This study looks at the application of mathematical concepts of entropy and Fibonacci sequence in creating optimal dimensional relations of classification character. The paper is devoted to optimization of some numerical relations and integers as unified threshold characteristics of classification type, aimed for example at systemic optimizing the measuring information of various processes. The paper is devoted to optimization of some numerical relations and integers as unified threshold characteristics of classification type. The dimensional concepts of information optimality, harmony and balance in the complex are major criteria underlying the present study. Based on the criteria, the maximum deviations from the optimal values and the way to optimize the redundancy of information are determined and justified. As the model in determining informational optimality has used the principle of equivalence, based on the equality of total entropy of system and the entropy of its informative components. The application of suggested theory is discussed on examples of characteristics relating to binary classification from the sphere of practical metrology, such as confidence levels, tolerance uncertainty ratios, and test uncertainty ratios. As accompanying results of the research the optimality of system of decimal numbering and the information nature of “magic number seven” are proved.