Optimizing architectural properties of Artificial Neural Network using proposed Artificial Bee Colony algorithm

The design of Artificial Neural Network (ANN) is a typical task as it is depends on human experience. There are few techniques like the Back-Propagation algorithm and nature inspired meta-heuristic are one of the most widely used and popular technique for optimizing feed forward neural network training. Artificial Bee Colony (ABC) algorithm is nature inspired meta-heuristic approach based on behavior of intelligent honey bees for searching food sources of its colony. To improve the performance of artificial bee colony algorithm, a novel proposed ABC approach is introduced using opposition-chaos initialization method with well balance characteristic among exploitation and exploration abilities. Finally the proposed novel ABC is used for optimizes an ANN´s architecture properties like synaptic weight and transfer function of each neuron that maximized the accuracy and minimizes the error. Analysis is been performed over standard classification dataset, reflecting light of efficiency of proposed method.