A MORPHOLOGICAL APPROACH TO SEGMENTATION OF MEDICAL DATA THROUGH CLUSTERING

Kraynov Kirill Alekseevich, Polyakov Evgeny Alekseevich, Kharchenko Elena Alekseevna

Typical problems specific to the segmentation of tabular medical data are highlighted (for example, the complexity of interpreting the results due to the need to modify the data). It is established that classical cluster analysis methods are often used to solve the problem of dividing medical data into homogeneous groups. An alternative method for clustering multidimensional data, isolated from the morphological method of making management decisions, is proposed. A method of its modification for working with large amounts of data is proposed. To confirm the effectiveness and convenience of the method, an example of processing one open near-medical dataset is given.

Keywords:system analysis, non-tagged data, structured data, mixed type data, big data.