Abstract

TOWARDS AN IMPROVED A NEURO-FUZZY CLASSIFIER FOR MEDICAL DIAGNOSIS

A.M. Sagir, (PhD)

---

Abstract. The aim of this research paper is to develop an improved a neuro-fuzzy classifier for the diagnosis of medical diseases. It further describes a methodology for developing the proposed classifier by incorporating the capability of fuzzy logic and artificial neural network learning algorithm that can be used by physicians to accelerate diagnosis process. The proposed classifier maximises the correctly classified data and minimise the number of incorrectly classified patterns. For robustness, the proposed method was tested with Diabetic Retinopathy (DR) and SPECT-Heart datasets obtained from University of California at Irvine?s (UCI) machine learning repository. In addition, an attempt was done to specify the effectiveness of the performance measuring classification accuracy, sensitivity and specificity. In comparison, the proposed method achieves superior performance when compared to conventional adaptive neuro fuzzy inference system (ANFIS) based gradient descent algorithm and some related existing methods. The software used for the implementation is MATLAB R2014a (version 8.3) and executed in PC Intel Pentium IV E7400 processor with 2.80 GHz speed and 2.0 GB of RAM. Keywords: ANFIS, Classification, Diagnosis, Medical disease data sets

PDF