You are on page 1of 2

Abstract

Development of any expert system requires a holistic understanding of methods


and their applications. An Expert System has five components viz; user interface,
explanation mechanism, inference engine, knowledge base and database of facts.
During the course of my study, I realize that the knowledge base is the most important
part of an expert system using artificial neural networks.
I initiated my work with a survey of existing research carried out between 2010
and 2015 in different countries pertaining to various areas like Medical, Education,
Automobile and Agriculture. I also conducted a literature survey on development of
Expert Systems using different methodologies and applications from 2005 to 2015.
I described the supervised learning for artificial neural networks, which is based
on the direct comparison between the actual output and the expected output. I
conducted a comparative study on various algorithms of supervised learning such as
decision tree, perceptron and back propagation.
I proposed methods to solve one of the common problems researchers may face at
the time of training of ANN which is over-fitting of outlier points. I clarified machine
learning and described the over-fitting and under-fitting as well as comparison
between them. I also described the methods for avoiding over-fitting and under-
fitting. I conducted a comparative study on various methods of machine learning to
avoid over-fitting and under-fitting.
I have also described different types of tools and found that the most ideal concept
is development of an expert system using tools which is a lot simpler as compared to
using programming language. Therefore, I conducted a comparative study between
MATLAB, Scilab, Prolog and Lisp using parameters like License, Cost, User
interface, Memory Requirement and Computation speed.
A NeuroFuzzy System (NFS) is one of the most commonly used systems in real
life problems, it has been recognized as a powerful tool which can facilitate the
effective development of models by combining information from different sources,
such as empirical models, heuristics and data. Neuro-fuzzy models describe systems
by means of fuzzy "if–then” rules represented in a network structure to which learning
algorithms known from the area of artificial neural networks can be applied. This
work examines a K-means clustering algorithm to enhance NFS efficient. The
i|Pa ge
enhancement of NFS by using K-means has been shown by running two of
experiments on nonlinear functions.
When fuzzy systems are highly nonlinear or include a large number of input
variables, the number of fuzzy rules constituting the underlying model is usually
large. Dealing with a large-size fuzzy model may face many practical problems in
terms of training time, ease of updating, generalizing ability and interpretability.
Multiple Fuzzy System (MFS) is one of the effective methods to reduce the number of
rules, increase the speed to obtain good results. This work is therefore proposes
another approach call Multiple Neuro-Fuzzy System (MNFS) which can further
enhance the performance of the MFS approach. The new approach is used Back-
propagation algorithm in the learning process. The performance of the proposed
approach evaluates and compares with MFS by three experiments on nonlinear
functions.
I proposed a new approach of developing of expert system knowledge base using
artificial neural network (ANN). It first defines Artificial Neural Network, discusses
an Overview of ANN-KB, suggests ways to insert Knowledge into a Neural Network
and tested it with two experiments. I proposed a method to develop the knowledge
base of expert systems using Artificial Neural Network (ANN). I designed Artificial
Neural Network (ANN) and tested it with perceptron and back propagation and also
as an standalone algorithm. The results showed that ANN when applied as an
standalone algorithm, it worked more efficiently and effectively as compared when it
is applied either with back propagation or with perceptron at least in extraction and
gathering knowledge. I also compared the current approach with that of the previous
similar works using two methods. I compared system that learn strictly from examples
and system that learn both from examples and theory.

ii | P a g e

You might also like