Identification of AIDS Disease Severity Using Genetic Algorithm - Computational Intelligence Techniques for Comparative Genomics - page 96

Permann [ 11 ] this paper describes initial research using GAs and helps to
optimize infrastructure protection and restoration decisions. This research proposes
to apply GAs to the problem of infrastructure modeling and analysis in order to
determine the optimum assets to restore or protect from attack or other disaster.

2 Dataset Description

Genetic research indicates that HIV originated in west-central Africa during the late
nineteenth or early twentieth century. HIV infection was identi ed in the early part

of the decade. Since its discovery, AIDS has caused an estimated 36 million deaths
(as of 2012).
The World Health Organization (WHO)

rst proposed a de nition for AIDS.

The WHO classi cation has been updated and expanded several times, with the

most recent version. The WHO Disease Control and Prevention also creates a
classi cation system that uses HIV infections based on CD4 count and clinical

symptoms of the infection in three stages:
Stage 1: CD4 count above 500 cells/mm 3 No AIDS Condition
Stage 2: CD4 count between 500 and 200. AIDS de

ning condition. (In this

condition to considered ART centers).
Stage 3: CD4 count 200 cells/mm 3 or AIDS de ≤ ning conditions. (Near Death

Stage).

Unknown: If insuf cient information is available to make any of the above

classi cations for surveillance purposes, the AIDS diagnosis still stands even if,

after treatment, the CD4+ T cell count rises to above 200 per mm 3 of blood or other
AIDS-de ning illnesses are cured (Table 1 ).

3 Methodology

3.1 Genetic Algorithms

GA is one of the ef cient optimization techniques which is based on “ Survival of

the cation rules

from the given set of training data items. The individual rule consists of

ttest. ” In this paper, we used GA to generate individual classi

“ if part ”

and “ then part. ” The “ if part ” is known as antecedent and “ then part ” is known as

consequent [ 9 , 13 ]. The rule representation is as follows:
If P then Q where
P is the list of attributes involved in the rule formation and
Q is the class label of the rule.