![](bg2.png)
2 CRITERIA OF DENGUE VIRUS
DISEASE
Dengue is the most common virus transmitted by
mosquitoes which causes up to 100 million
infections and 25,000 deaths worldwide each year.
2.1 WHO Criteria
WHO announced a set of criteria for classifying
dengue patients according to DF and DHF (see
Table 1 for details).
However, WHO criteria are not sufficient to
classify the dengue patients. Since they are just a
common criteria for dengue virus disease. We
believe that there are some different clinical
evidence and laboratory results that fit to our
regional disease. There are some researchers work in
this area such as (Tanner, et al, 2008) and (Tarig, et
al., 2010). They tried to find new criteria in order to
classify dengue patients.
2.2 Tanner’s Criteria
(Tanner, et al, 2008) employed decision tree to
classify data into 4 levels which are Probable
dengue, Likely dengue, Likely non-dengue and
Probable non-dengue. Their data set contains 1,200
patients (1,012 patients from the EDEN study and
188 patients from Vietnam). They found 6
significant features which were platelet count (PLT),
white blood cell count (WBC), body temperature
(T), hematocrit (HCT), absolute number of
lymphocytes (Lymphocyte) and absolute number of
neutrophils (Neutrophil). They got 84.7%
correctness.
2.3 Tarig’s Criteria
Research work done by Tarig Faisal (Tarig, et al.,
2010). Showed that, they can predicted the risk of
dengue patients using Self Organizing Map (SOM)
and Multilayer Feed-forward Neural Networks
(MFNN). Nevertheless, their accuracy rate was only
70 %. Their next research was to do data clustering
on patients into 2 groups as low risk and high risk
patients. They classified 195 patients using three
criteria obtained from SOM. There are 3 risk criteria
which were platelet counts (PLT) (less or equal than
40,000 cell per mm
3
, HCT (greater than or equal to
25%) and aspartate aminotransferase (AST) (rose by
fivefold the normal upper limit for AST or alanine
aminotransferase) (ALT) (rose by fivefold, the
normal upper limit for ALT) A high risk patient was
a patient who had at least 2 criteria. A low risk
patient was a patient who had less than 2 criteria.
Their finding supported the criteria of WHO. Lastly,
in June 2010, they classified the risk of dengue
patients using MLP. The accuracy only 75 %.
(Ibrahim et al.,2005) predicted the day of
defervescence of fever (day0). Their data set
consists of 252 dengue patients (4 DF and 248
DHF). They applied Multi-Layer Perceptron (MLP)
and got 90 % correctness.
3 DATA PROCESSING
Data integration is a step that integrated the data
from several sources. In this study, Siriraj Hospital
integrated patient’s data from Srinagarindra Hospital
and Songklanagarind Hospital. Next step is data
cleaning. Sometimes the data sets contained noise
data that results from human error or machine error.
Table 1: WHO criteria (World Health Organization, 1999).
Symptoms Laboratory
DF
Fever with two or more of the following
signs: headache, retro-orbital pain, myalgia,
arthralgia.
Leukopenia occasionally. Thrombocytopenia,
may be present, no evidence of plasma loss.
DHF I
Above signs plus positive tourniquet test Thrombocytopenia < 100,000, HCT rise >=20 %
DHF II
Above signs plus spontaneous bleeding Thrombocytopenia < 100,000, HCT rise >=20 %
DHF III
Above signs plus circulatory failure ( weak
pulse, hypotension, restlessness)
Thrombocytopenia < 100,000, HCT rise >=20 %
DHF IV
Profound shock with undetectable blood
pressure and pulse.
Thrombocytopenia < 100,000, HCT rise >=20 %
DATA MINING ON DENGUE VIRUS DISEASE
33