Final Paper Fuzzy Intelligence Time Allowed 3 Hours
Consider a class C of geometrical curves that could be characterized by 4 continuous numeric
parameters named p1, p2, p3 and p4 with normalized values in the range of 0 to 5. Given a curve c in
the class C, the aim to put it in one of 10 subclasses (C1, C2, C3, .........., C10) of C. A statistical
research on these curves shows the following.
a. The curves belonging to C1 have:
i. Normally distributed value of parameter p1 with mean 2 and variance 1, AND
ii. Normally distributed value of parameter p2 with mean 2 and variance 1.
b. The curves belonging to C2 have:
i. Normally distributed value of parameter p1 with mean 3 and variance 1, AND
ii. Normally distributed value of parameter p3 with mean 2 and variance 1.
c. The curves belonging to C3 have:
i. Normally distributed value of parameter p2 with mean 3 and variance 1.
ii. Normally distributed value of parameter p3 with mean 3 and variance 1.
iii. And parameter p1 is absent.
Similarly other classes are also investigated in this research.
Use this information to design a fuzzy classification system to find which class a particular given
curve belongs to. Assume there is a features extraction system that gives the parameter values p1,
p2, p3 and p4 for a given curve. You can use appropriate triangular membership functions and
consider only three classed described above. Your system is expected to take the values of
parameters p1, p2, p3 and p4 as inputs and return membership values of a given curve for the above
three classes (please provide all the steps of the algorithm to perform this classification and explain
with diagram where required). (30 Marks)
Compared to an Artificial Neural Network (ANN) classifier used for the same purpose, how the fuzzy
classifier could be advantageous? (10 Marks)
Compared to ANN classifier, what are limitations or disadvantages of this classification? (10 Marks)
docsity.com
