This article is dedicated to the scientific visualization method for the task of big data analysis. With the help of well proved methods of visual analytics the approach undertaken was directed at the criterial function optimization in medical diagnostics problems. The visual method being used makes it possible to represent error function and analyze its behavior in 2D and 3D, solving the problem of optimal classification of two sets. The visualization method made it possible to reveal a variety of error function characteristics which were obtained by authors as by a practical results generalization while processing medical data bases. An example of the effective use of visualization method is the application to optimize the diagnostics of breast cancer patients. At the stage of diagnostics, the possibility of optimizing by pre-determination the structural characteristics of the multidimensional structures in the parameter space of the linear discriminant function was investigated. As the criterion function, a discrete error function was used that registers the number of incorrectly recognized objects. Some currently used classification techniques are poorly adapted to find the minimum of medical diagnostics problems and are time- and resource-consuming. Others give a significant probability of error, that is unacceptable in the diagnostics of a particularly responsible decision-making area. Therefore, the authors proposed a simple and visually intuitive method to determine the structural characteristics of the error function. We used a visual method of investigation consisting of the algorithm of constructing the equal level surfaces and used them to go down to a minimum. As a result, it was shown that a sufficiently reliable diagnosis can be made, if we apply the method for visualizing the discrete error function when running on extremal to a minimum. With the help of a visualization method the software has been developed, as a result of which we get an accuracy of 94.73% when a benign tumour was separated from a malignant tumour using two diagnostic features.