Statistical modelling with neural network and an attempt for an application

11 ABSTRACT Master of Science Thesis STATISTICAL MODELLING WITH NEURAL NETWORKS AND AN ATTEMPT FOR AN APPLICATON Halil ERYILMAZ Anadolu University Graduate School of Natural and Applied Sciences Statistics Program Supervisor: Prof. Ali Fuat YÜZER 2004, 104 pages Imitiating the principles of human brain, Neural Networks can produce meaningful answers for never seen or never encountered problems via creating generalizations from the present information for the relationships (linear or not) between input and output. In some sources published early 1990's, it's brought forward that there were some resembles between some neural network models and statistical techniques and even for some cases there was equality. Later studies showed that it was not just pure coincidince but these two scientific areas were related to each other extensively. Comparison neural networks and statistical techniques showed that either is useful for creating better results on the other. A few scientist put forward that some of the neural network models like perceptron or multilayer perceptron were actually very useful for statistical applications. The aim of this study is to explain neural networks, to show the relationships between neural network models and statistical techniques and using a real life application to show efficiency and applicability of neural networks in statistical problems. Keywords: Neural Network, Perceptron, Multilayer Perceptron, Backpropagation Net, Logistic Regression