Modern visualization methods are used to convey information about an object or process and as a tool for search and decision-making process. Data and signals, in AQ1 analog and digital form, are only valuable if they are analyzed for a specific goal. In this work we etablish the classification of visualization tasks from the point of analyzing heterogeneous multidimensional data, including the AQ2 case when at the initial stage it is required to formulate a research hypothesis. A classification of visualization metaphors is presented, which is necessary for a conscious choice of tools for visualization and data analysis. This is important for understanding and managing the interpretability of information, the formation of the correct meaning and operational understanding. We demonstrate examples of static and dynamic models of visualization. Based on the semantic model and proposed classification, the principles of visual metaphors formation for solving several applied tasks in various fields of knowledge (oil and gas production, biomedicine, materials science, education, management, etc.) are formulated.