Dekompresyon hastalığının cluster analizi

There have been many classifications of Decompression Illness (DCI) which is seen in divers as the result of bubbles which expand in human body causing local damage in tissues or which block blood circulation because of decompression.The diagnosis and classification of DCI is made observing the patient?s symptoms and signs. The treatment is performed in a hyperbaric chamber where the conditions are reversed (recompression) and the combination of pressure and time is determined by the type of the disease.The problem is that DCI has a lot of signs and symptoms, resulting in a lot of different classifications of the illness requiring different treatment plans which makes the correct classification of DCI extremely important and data mining techniques can be used as decision support tools to determine the type of DCI.In this thesis we classified empirically the DCI patients using the sign and symptom list of the Diving Injury Reporting Forms (DIRF) of Divers Alert Network with different clustering algorithms (k-means, COBWEB, EM) and compared our results with recent statistical studies on DCI classification and other classifications and outcome of treatment. And we have also found association rules which will contribute differential diagnosis.Consequently, the classes we have obtained after clustering have the characteristics of hierarchy from mild to severe as in other classifications and as in recent classifications of DCI.Keywords: Decompression Illness, clustering, association rules, data mining, association, COBWEB, EM, K-Means