Breast Cancer Diagnostic Factors Elimination via Evolutionary Neural Network Pruning

The redundant medical diagnostic factors elimination problem was faced by the use of Artificial Neural Networks that were evolved using Genetic Algorithms. For specific medical diagnosis problems such as breast cancer classification, with given set of diagnostic input parameters, Genetic Algorithms were used for pruning Neural Network structure and the investigation of the most appropriate subset of input parameters of Artificial Neural Networks that can still provide reliable medical diagnosis. Neural Networks were pruned in both the input as well as the hidden layer(s) by Genetic Algorithms that were utilized to search for pruned Neural Networks with the same, or even improved performance and therefore with enhanced medical classification and diagnostic ability of the original full-sized Neural Networks. Neural Network pruning and size reduction without loss of diagnostic ability can support redundant medical diagnostic factors or parameters elimination. Key-Words: Artificial Neural Networks Pruning, Genetic Algorithm Optimization, Breast Cancer Classification