Conventional Neural Network approach for the Diagnosis of Lung Tumor

Abstract

The aim of this research is to develop an Optimal Classifier based on computational intelligence techniques for the precise diagnosis of deadly Lung Cancer disease. The proposed system provides maximum classification accuracy along with minimum number of connection weights, processing elements, time elapsed per epoch per exemplar and MSE on CV data sets. The Classifiers based on MLP, GFF, MNN Neural Networks and SVM with different learning rules on different transform domains such as DCT, FFT and WHT have been simulated on two different datasets. The optimized single hidden layer Multilayer Perceptron Neural Network with QP learning rule on Histogram knowledge-base for Data-base I and Data-base II resulted into the reasonable and optimal classifier based on C.I. techniques for the diagnosis of Lung Cancer.