The main goal of this case study is to develop an artificial intelligence system using neural networks to classify the diagnosis of breast cancer into two types of cases based on the selected characteristics which are malignant (cancer cell) and benign (normal cell). The performance of Multilayer network using three different types of activation function will be analyses and compared to produce the best network with the highest probability of convergence, fast training and, powerful network. The scope of this project is to investigate and studies a developed system that applied a neural network method which combines processes of selecting activation functions to obtain a desired output. In addition, the scope of this project is also to develop Multilayer Perceptron network using Borland C++ Builder and MATLAB Neural Network Toolbox. This project focuses more on designing and writing code in Borland C++ Builder environment. MLP network with Back Propagation learning algorithm and will be developed to identify the suitable and powerful activation function that can produce the best output node and can make the training much faster. The various type of activation also has been studied and investigated using MATLAB. The developed Artificial Intelligent system will be used to detect and classify breast cancer cells which are malignant cell and benign cell. Beside the main purpose there are several goals that need to achieve which are run a test and make an analysis towards system that has been developed to yield a high performance and accuracy of breast cancer result and identify the suitable learning
algorithm and powerful technique that can obtain the powerful approach to produce the best network with highest accuracy of breast cancer possibilities.
Effect of activation function in mlp network and its variety_Norhaslinda Aziz_E3_2007_NI