Application of OWA Based Classifier Fusion in Diagnosis and Treatment offering for Female Urinary Incontinence

摘要

Classifier fusion is a process that combines a set of outputs from multiple classifiers in order to achieve a more reliable and complete decision. In this work, the application of Ordered Weighted Averaging (OWA) operator as a classifier fusion approach, for diagnosing and offering the treatment of female urinary incontinence has been investigated. In this study, a classifier combination system has been constructed on four underlying individual classifiers, with different approaches including two multi-layer perceptrons, a generalized feed forward and a support vector machine. The system combines the decisions of these classifiers and is considered as a medical council based on only clinical patients data. Instead of choosing very accurate and expensive data sources like urodynamic, cystoscopy and voiding cystourethrogeram as paraclinical tests, we can nominate a small group of experts and use not so costly clinical measurements and then take experts' judgments and weight them by the level of expertise they have. Considering only clinical patient data which gathered from Iran urology medical center, the accuracy of OWA based classifier fusion system in diagnosis of urinary incontinence types improved 2.02%, 4.11% and 8.27% comparing the accuracy obtained by best individual underlying classifier, simple averaging and majority voting respectively.