Sağlık Hizmetlerinde Aykırı Dataların Kestirimi İçin Mekansal Zamansal Veri Madenciliğinin Kullanımı

Authors : Abdulsamet Haşiloğlu, Seyma Yücel Altay, Umit Ertaş

Pages : 411-428

Doi:10.31202/ecjse.315044

View : 16 | Download : 13

Publication Date : 2017-09-30

Article Type : Research

Abstract :In the last decade, useful information extraction from moving objects has become widespread in the spatial-temporal data mining field with the increasing use of devices such as RFID and GPS. For this purpose, the outlier detection method, which is a subfield of data mining, was applied to the trajectory of patients and diseases in the dental health service. In this article, TRAOD and TOD-SS algorithms combining distance and density-based methods were preferred. These algorithms do not handle the moving object trajectory as a whole unlike other outlier detection techniques. They investigate whether each piece exhibits different behavior according to its neighbors by separating trajectories into pieces. So, they detect outlying trajectory pieces that other algorithms cannot locate. Algorithms preferred in this study were used in a COMB-O model we developed and their performances were compared. In addition, according to the region and clinic, the classification of patients was made. Also, clustering, which is another branch of spatial-temporal data mining, was performed for trajectory. When the COMB-O model was executed, results showed sub-trajectories that deviated from the trajectory data were successfully detected with the help of the trajectory outlier detection algorithms. Inconsistent trajectories perceived provided significant data. In addition to this, successful classification was performed by making use of non-linear classification features of DVM. Moreover, stops and moves in the Faculty of Dentistry were detected by using CB-SMoT and DB-SMoT which are clustering algorithms.
Keywords : Spatio-temporal veri madenciliği, hareketli cisimler, yörünge verileri, yörünge sapması algılama, destek vektör makinesi