Design and analysis of a multivariate regression model using artificial neural network

Muhammad Rizman , Mohd (2014) Design and analysis of a multivariate regression model using artificial neural network. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text DESIGN AND ANALYSIS OF A MULTIVARIATE REGRESSION MODEL USING ARTIFICIAL NEURAL NETWORK 24pages.pdf Download (331kB)

Abstract

Neural network provide an application that can be applied in a broad range. It is a great new technique for solving problems in many different disciplines. The purposed of this study is to develop a model of artificial neural network based on a model of regression. In a complex system, using hard computational such as regression method could cause several problems to occur and the accuracy of the system cannot be inferred. In order to study the performance of neural network, an experiment is designed. High speed end milling is one of the manufacturing processes that are selected for the experiment conducted based on the design of experiments using Box-Behken method. 29 samples were conducted using variables parameters such as cutting speed, feed rate and depth of cut. The data of surface roughness obtain are collected to be used in analyzing the model of neural network and multiple regression. Regression model using the second-order form while neural network using fuzzy cognitive network algorithm with differential hebbian learning rules and linear activation function. The mathematical model developed using multiple regression method demonstrate the accuracy of 96.637%, while as for the model developed using neural network exhibit an accuracy of 99.999976% and 99.9999596% for training and testing stage respectively, in predicted of the surface roughness. This shows the feasibility and applicable of neural network compare to multiple regressions. This result is as a guide for future research or implement on the complex system.

Actions (login required)

View Item