Skip to Main Content
This paper compared various MLP activation functions for classification problems. The most well-known (Artificial Neural Network) ANN architecture is the Multilayer Perceptron (MLP) network which is widely used for solving problems related to data classifications. Selection of the activation functions in the MLP network plays an essential role on the network performance. A lot of studies have been conducted by reseachers to investigate special activation function to solve different kind of problems. Therefore, this paper intends to investigate the activation functions in MLP networks in terms of the accuracy performances. The activation functions under investigation are sigmoid, hyperbolic tangent, neuronal, logarithmic, sinusoidal and exponential. Medical diagnosis data from two case studies, thyroid disease classification and breast cancer classification, have been used to test the performance of the MLP network. The MLP networks are trained using Back Propagation learning algorithm. The performance of the MLP networks are calculated based on the percentage of correct classification. The results show that the hyperbolic tangent function in MLP network had the capability to produce the highest accuracy for classifying breast cancer data. Meanwhile, for thyroid disease classification, neuronal function is the most suitable function that performed the highest accuracy in MLP network.