International Science and Technology Journal

This paper presents Multi-class pattern classification using neural networks. Classification is the process of mapping a vector of observed characteristics into a defined class. It can be separated into two classes, which is binary classification andmultiple classes which is Multi classification. In binary classification, only two classes are involved while, in multiclass classification, there are more than two defined classes. Most of real-world classification problems has more than two classes, such as text document classification, speech recognition, object recognition…etc. Dealing with large number of classes can be problematic, even for small number of classes some techniques may become prohibitive to use in practice. We will discuss different approach to multi-class classification, that reduces the problem to that of multiple binary classification tasks, and then we will focuses on issues including neural network architecture in classification. Neural Networks considered one of the most successful approaches for multi-class classification problems, where it maps the input feature vector to the network output that contains more than two classes "K-class pattern". This algorithm is to train the hidden neurons separately to classify all the different classes, where the numbers of neurons determine the number of classes in the classification