A hybrid intelligent system for data analysis and visualization
Loading...
Date
2007
Authors
Wang, Shir Li
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In this research, a hybrid system consisting of the multilayer perceptron (MLP)
neural network and the circle-segments method for data analysis and visualization is
designed and developed. Acting as a black box, the MLP network normally gives a
prediction without providing a facility for users to visualize the solution. As such, this
research proposes to hybrid the circle-segments method with the MLP network,
whereby the circle-segments method is used in two different ways, i.e., to provide visual
correlation between the input-output data samples to users and, thus, to allow users to
eliminate insignificant inputs from the input data set.
The effectiveness of the proposed MLP-circle segments system is evaluated
using a number of benchmark and real case studies. For process modelling and
prediction problems, the case studies investigated include the Friedman#1 benchmark
problem, wire electrical discharge machining (EDM) process, disk drive read control
system, and PID controller tuning. For data classification problems, the case studies
investigated include the Iris, and Wine benchmark problems, as well as a real medical
problem pertaining to acute stroke diagnosis.
In process modelling and prediction problems, the performances of the hybrid
system are compared with those from the response surface methodology (RSM). The
proposed system achieves an improvement of at least 11% in term of accuracy as
compared with the results from the RSM. In data classification problems, the results
are compared with those from MLP coupled with the principal component analysis
(PCA) as well as MLP without any feature selection method. It is found that the
accuracy of proposed system is as good as, if not better than, MLP coupled with the
PCA and MLP without any feature selection method. Based on the results obtained, the
proposed MLP-circle segments system demonstrates better prediction and visualization
abilities, thus justifying its potentials as a useful and effective system for data analysis
and visualization. The proposed MLP-circle segments system is useful in data analysis
and visualization, especially in the domain of process modelling and prediction, as well
as data classification problems.
Description
Master
Keywords
Computer Science , Hybrid intelligent system , Data analysis