Artificial Neural Network Approach To Certain Classification Problems (Paperback)

This book focuses on Artificial Neural Network (ANN) based statistical clustering technique to group the objects and improve the classification methodology. The self-organizing principle and the neighbourhood concept of ANN in Kohonen's Self-Organizing Map (SOM) are used develop a new statistical clustering technique. The proposed clustering algorithm uses statistical distance measure and the concept of neighbourhood of input vector. This algorithm is tested using some real and simulated data sets, and it has yielded better results. Textile data and finance data have used to illustrate real time applications of the proposed clustering technique. The cluster analysis can also be performed on the non-statistical data by introducing Euclidean distance measure or any such distance measure. Repeated discriminant analysis suggested in conjunction with the ANN based clustering method will be very useful for data satisfying statistical requirements.