Genetic Algorithm-Artificial Neural Network (GA-ANN) Hybrid Intelligence for Cancer Diagnosis

Abstract

Artificial Neural Network (ANN) is one of the most promising biological inspired computational intelligence techniques. However designing an ANN is a difficult task as it requires setting of ANN structure and tuning of some complex parameter. On the other hand, Genetic Algorithm (GA) as a global search technique is useful for complex optimization problem where the numbers of parameters are large and difficult to obtain. In this paper GA has been used to simultaneously select significant features as input to ANN and automatically determine the optimal number of hidden node. Meanwhile the ANN training is done by Levenberg Marquardt (LM) algorithm. A new procedure in obtaining optimal ANN architecture is also described which based on feature importance determine by Genetic Algorithm. Simulation results on cancer dataset proved that the proposed method has achieved the highest 97% average percentage of correct classification with the absent of 2nd and 5th feature.