Designing Multi-layered Articial Neural Networks for Risk Analysis of Lung Cancer Disease

The paper is intended to propose a multi-layered artificial neural network model for lung cancer risk analysis. Hundreds of thousands of people die every year due to cancer; use of mathematical models especially in decision making may help reduce the loss of patients by augmenting effective predictions in early stages of diagnosis. Feed forward back propagation and cascade forward back propagation network structures have been applied for creating multi-layered ANN model. Levenberg-Marquardt algorithm and Bayes regulation algorithm have been used for training step. Hyperbolic tangent sigmoid function has been chosen as the transfer function and the learning algorithms have been integrated. Four multi-layered ANN models are evaluated with regression and validation analysis on a dataset of 616 people. A comparative analysis is also performed to show that our bayes1ffb model gets higher accuracy value than two existing models in the literature.