The paper proposes a new approach to the agglomeration of data in cluster analysis. The new approach assumes that sets of similar events are attributed to the cumulative probability of their occurrence at the same time. Such approaches will not be found in probability. Thanks to the mathematical theory of records fairly accurate classification of the object can be provided. This is the method which can be used in the cluster analysis by agglomeration. Figure 1 has been drawn for the purposes of better illustration of the problem. It shows the problem of classifying an object to one of the two classes: suitable or unsuitable for further use. Thanks to the merger of two classifiers: KNN algorithm (k nearest neighbours) and belief function a model was created, which is pretty strong as it seems to discriminate against space objects. It therefore seems reasonable to discriminate space of objects. The paper also shows a possibility of applying the proposed model to classification and the correlation between cytokines and features related to the occurrence of lymphocytic leukaemia. It therefore seems justified to carry out tests on this new method as regards various scientific problems.
Computer recognition of data structures using cluster analysis and the theory of mathematical records
2017
Aufsatz (Zeitschrift)
Elektronische Ressource
Unbekannt
Metadata by DOAJ is licensed under CC BY-SA 1.0
Voice-Recognition System Records Inspection Data
NTRS | 1993
|Mathematical Analysis and Speaker-Independent Speech Recognition
British Library Online Contents | 1996
|Computer storage of accident records
Engineering Index Backfile | 1967
|Structure Analysis and Recognition of Mathematical Expressions
British Library Conference Proceedings | 1995
|