Kohonen Öz Örgütlemeli Haritalama Yöntemi ile Psikotik Hastalıkların Kümelenmesi
xmlui.mirage2.itemSummaryView.MetaDataShow full item record
Data mining is a method that is used to extract hidden, valuable, usable information among large amounts of data, and to support strategic decisions. Despite the fact that data mining is used in a variety of fields, it faces various challenges, especially in the field of health. The constant change of information in the health field, the excessive amount of non-structural data, and the presence of these data in different environments make it difficult to analyze data. It is thought that classification or clustering by 'Kohonen self-organization mapping method' in cases where classification or clustering problem is experienced will solve classification and clustering problem. The aim of this thesis is to theoretically introduce the 'Kohonen self-organization mapping method' used in data mining for classification and clustering purposes, to examine its properties, to classify/cluster the subtypes of psychotic diseases with related methods, to show its applicability of this method in case of necessity of classifying or clustering in health field. The thesis hypothesis is that the classification / clustering performance (accuracy rate, etc.) obtained as a result of this method will be higher than the other classical methods. In this thesis study, data of 268 psychotic individuals collected for the dissertation thesis named 'Factor structure in schizophrenia and other psychotic disorders' were used. Clustering and classification methods were analyzed using the R program. When the performance of the classification made by the Kohonen self-organization mapping method was evaluated, it was found that the classification performance was high. The clustering performance of the Kohonen self-organization mapping method on health data was investigated and the result of the Kohonen self-organization mapping method was compared with classical clustering methods (k-means method and hierarchical clustering method). Clustering performances were evaluated using clustering indexes. In terms of 4 different diagnostic groups, the results of Kohonen self-organization mapping method and other clustering methods were compared. It was seen that the success rate changes according to the diagnostic groups and indexes, and only in one index Kohonen self-organisation mapping map has better performance than the other methods.