ISSN: 2455-8826

Asian Journal of Innovative Research in Science, Engineering and Technology

(AJIRSET)
Available online at: www.engineeringjournal.info

RESEARCH ARTICLE

An Intelligent Medical Decision Support System

Ogbonnaya S. Success, LN Onyejegbu*

University of Port-Harcourt, Rivers State, Nigeria.

*Corresponding Author: Email: nneka2k@yahoo.com

Abstract

Continuous interpretations of numerous health examination results of patients are tedious and
monotonous for clinicians; that are more often than not burdened by the overwhelming and seemingly
intimidating number of patients they need to attend to routinely. This has culminated in a lot of fatal
errors on the part of the medical practitioners which has led to the loss of innocent lives. Medical
diagnosis is an important but complicated task that should be performed accurately and efficiently and
its automation would be very useful. In this work, we developed an intelligent medical decision support
system that would aid medical personnel in the accurate diagnosis and prediction of hepatitis A and B
using C5.0 algorithms. A comparison was made between using the proposed C 5.0 Algorithm, and
existing C4.5 algorithm. The simulation result shows that the Mean Absolute Error for C5.0 is 0.04 while
that of C4.5 is 0.09%. The time complexity of the process for C5.0 is 0.03 seconds while that of C4.5 is
0.05 seconds. The correctly classified instance of C 5.0 is 98.05% while that of C4.5 is 94.81%. The result
shows that C5.0 algorithm performs better than C4.5, in the diagnosis and prediction of hepatitis A and
B. C5.0 machine learning algorithms was used to train the historical datasets of the hepatitis virus.
WEKA simulation software was used to compute and simulate C5.0 algorithm. The methodology adopted
for this work, is Object Oriented Analysis and Design Methodology. PHP programming language, HTML
and MySQL database was used for the implementation.

Keywords: Hepatitis, C4.5 and C5.0 Algorithm, WEKA, Decision Tree, Medical decision Support system.

Introduction
Health examinations produces full and
detailed information about the health
condition of patients, the obtained
information is needed for the management of
individual and public issues of health.
Examining health conditions does not only
give health practitioners needed information
to aid early diagnosis of diseases but also
supplies the health examination receivers’
proper recommendation for their self-health
management. Both health practitioners and
receivers of health examination derive
benefit from the comprehensive information
supplied, to obtain meaningful result from
the examination process.
Brause [5] opined that most health
practitioners are posed with lots of task one
of which is learning to diagnose in their early
year of study. For them, the highest task in
health evaluation is generating a
comprehensive report after interpreting the
result gotten from the examination. The
results gathered from distributed sources of
test are simultaneously and entirely
interpreted to get the actual picture of health
conditions of every health evaluation
receiver. However, the detailed
interpretation of multiple examination items
is complex and error-prone for a busy
clinician.

A detailed health examination package may Continuous interpretations of numerous

compose of several examination items such as health evaluation results of patients are
one on one physical examination, tedious and monotonous for health
endoscopies, laboratory experiments and test, practitioners. Health practitioners have to
radiological studies, and others. The task of spend much effort in screening plenty of
health science is to diagnose and prevent health evaluation results to detect relatively
disease; the focus here is on diagnosis. few clinically significant findings. It is most

LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12 1

 Available online at: www.engineeringjournal.info
likely to have a reduction in the number and
value in terms of quality of health evaluation
result interpretations at the situation of
human fatigue. Under this circumstance,
health practitioners tend to simplify the
procedures such as following their own
experience rather than clinical guidelines to
interpret clinical data and make decision. All
these conditions potentially affect the most
important issues that concerns health
evaluations and quality control of all the
health examination reports.
In a need of obtaining a practical, more
accurate and effective computer-assistance
for health evaluations impels the
development process of the medical decision
support system (MDSS) proposed in our
research. This research developed an
intelligent model for medical decision support
system (MDSS) that would aid human
medical personnel in the accurate diagnosis
and proper prediction of Hepatitis cases A
and B.
This was be achieved by developing an
intelligent rules from hepatitis data set using
SEE5.0/C5.0 decision tree algorithm, a model
for intelligent medical decision support
system (DSS) was developed using the
developed rules, the system was
implemented using PHP, HTML and MySQL
as the database. The data for this research
was gotten from University of California
Irvine (UCI) Machine Learning Repository
database.
Review of Related Work
The literature review measures the extent to
which work has been done in relation to
intelligent medical decision support system.
Evan DA [9] describe four various levels
whereby medical knowledge are structured in
a clinical problem solving framework. They
identified that observations are entities of
evidence that are known as possibly
pertinent in a difficult solving situation. No
useful clinically facts were constituted.
Facets reflect general pathological
descriptions such as aortic insufficiency, or
categorical descriptions such as endocrine
problem. They are also interim hypotheses
that divide the information in the problem
into manageable sub-problems and suggest
possible solutions. Facets vary in their level
of abstraction - from high level facets which
LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12
may partition the problem space and may
likely be a meaningful approximation to a
candidate solution - to low level facets which
may involve a more local inference that may
explain one-or two findings and would no
advanced the problem solving process to the
same extent.
Oguntimilehin A et. al. [14] discoursed that
clinical diagnosis is basically categorization
which permits physician to forecast using
medical conditions and to decide suitable
cause. This is a difficult decision procedure
that encompasses large volume of ambiguity
and also uncertainty organization mostly
when dealing with disease of multiple
symptoms. This experience builds a system
that powerfully maps the connections,
observations and corresponding diagnoses.
Dragulescuand Albu [8] implemented a
system that is made of three important parts.
First of logical inference is used to decide
what type of Hepatitis virus is present for a
new patient. The possibilities are B, B+D and
C. After the second portion of the program
was employed to understand what will the
type be and the grade of hepatitis B (if the
sick person is diseased with the virus,
hepatitis B). This branch of the application is
developed using methods from statistical
inference. The third\ section of the
application consist of the patients ill with C
virus and it predicts the biological
parameter; evolution during the treatment
using AI Network.
Kahn R. et. al. [11] proposed neutrosophic
classifier which employs Neutrosophic logic
and is an addition of normally fuzzy classifier
used. They compared it using the resulting
parameters: like sum rules and the nature of
associated functions. They proved that
lengthy fuzzy classifier: Neutrosophic
classifier; optimizes the believed parameters
in comparison to fuzzy equivalent. Finally
they established that Neutrosophic logic still
in its emerging stage quiet holds the probable
to be further experimented using different
fields.
Ahmed [1] used neutrosophic and fuzzy C/
means gathering for enhancing CT liver
appearance. Abdominal CT imaging was
chosen and a segmentation approach was
given to get their capability and accuracy.
2
 Available online at: www.engineeringjournal.info
The image is transmuted into Neutrosophic
domain, this is designated with three
subsets. The ratio of fact in the subset T,
ratio of indeterminacy using subset 1, the
ratio of inaccuracy using subset F. threshold
of subset T, I, F was adapted using fuzzy C
mean algorithm.
He concluded that Neutrosophy can reduce
over segmentation and gives a better
presentation on deafening and irregular
images gotten using other approach since the
technic can handle ambiguity and
indeterminacy better. Ming R et. al. [12]
applied neutrosophy to image processing by
defining a-neutrosophic domain, which is
termed by three subsets T, 1, and F. They
then employ watershed algorithm to perform
segmentation of the shape in the
ncutrosophic domain. The experiments
showed that the plan method got better
results when compared with the existing
method.
Anil [3] in their, work titled "A soft
computing paradigm for a medical data
mining tool to predict risk of coronary heart
events", they proposed a neuro-fuzzy soft
computing data mining tool which was used
for the prediction of the severity of risk in
coronary heart events. The proposed system
was based on medical data obtained from
clinical observations performed on cardiac
patients.
They used the concept of decision tree
founded on 1D3 algorithm to classify the
feature that contributes towards the
diagnosis. The outcome is transformed into a
crisp if then rules and then transformed into
fuzzy rule base. A neuro fuzzy method is
given to improve the fuzzy membership
function which/ outcomes in the improvement
of the old system. The efficiency of the newly
generated expert system, prediction accuracy
is presented against the real life which was
established to be operational.
Djain and Kimbi [7] proposed a medical
diagnostic support system for the
management of hypertension (MEDIAG).
Their proposed system diagnoses the
possibility of the ailment and its severity by
applying fuzzy logic approach. MEDIAG
applied forward chaining method in making
inferences and the root sum squares of
LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12
drawing inferences was employed to infer the
data from the rules developed. The/ de-
fuzzification method engaged is centroid
approach. The result obtained resembles
human decision making with its capability to
work from approximate reasoning and
ultimately find a precise solution. It has 85%
exact diagnosis.
Nicolae V et. al. [13] presented an approach
that was ment to integrate computational
intelligence with clinical investigation
methods and knowledge for glaucoma
diagnosis. The knowledge acquired is
embedded in a fuzzy logic inference system.
The resulting neuro fuzzy glaucoma
diagnosis and prediction system is expected
to lower the effort, difficulties and risk cost to
this disease. Voula and [6] presented a new
hybrid modeling methodology suitable for
complex decision making procedures. Their
work extended previous work on competitive
fuzzy cognition maps for a medical decision
support system by complementing them with
case based reasoning methods.
The synergy of these methodologies is
accomplished by the planned decision
support application that is suitable to handle
situations where decisions are not clearly
distinct. The methodology developed was
applied successfully to model and test two
decision support systems, one a differential
diagnosis problem from speech pathology
area for diagnosis of language impairments
and the other decision making choices in
external beam radiation therapy.
Chrysostomos [10] presented an expert
system for differential diagnosiserythemato-
squamous disease incorporating decisions
made by three classification algorithm:
nearest neighbour classifier, naive Bayesian
classifier and voting feature interval 5. Their
proposed system enables doctors to
differentiate six types Erythemato-squamous
diseases using clinical and histopathological
parameters obtained from a patient-The
systein also gave explanation for the
classification of each classifier.
Gt Venira [4] presented a comparison
between Backpropagation and Naive Baye
Classifiers to diagnose Hepatitis disease. The
performance of proposed methods was
selected for each classification tasks of
3
 Available online at: www.engineeringjournal.info
hepatitis disease. The overall accuracy of
diagnosis systems were 98% and 97%
respectively.
Sonawane [15], system ascertained that
program based Medical Decision Support
System (MDSS) will be very helpful for the
medical professional due to its accurate and
speedy decision making. They discoursed
that forecasting the presence of heart ailment
accurately, effects in lives saving. Sequential
least optimization technique was use in
support of vector machine (SVM). Database
that exists at the Cleveland heart disease
gotten from UCI repository was trained using
SMO methods. The result shown made us to
know that the MDSS is able to diagnosis
faster the heart disease and produce a more
correct result; it was also show excellent
capability to predict large dataset.
According to Amin SU [2] medical errors are
harmful and costly. It leads to thousands of
mortality yearly. Hence, CDSS would give
chances to decrease medical errors so as to
enhanced patient safety. The proponents
started a reasonable examination of the
behaviour and performance of the six CDSS
programs having diverse data mining
methods for the diagnosis of heart disease. It
was concluded based on results that the
system cannot identify treatment or
management options for patients with heart
disease.
Sonawane [15] emphasized that decision
methods integral in operating theatre body
are a lot subjected to investigation, which
occasionally indicate a far results from
optimal. They additional uphold that the
pending surgery lists had constantly been a
general problem, partly owing to the point
Figure 1: overall prototype Of CDSSs (Source: Lincoln 2001)
LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12
that the present hospital systems existing in
Portuguese hospitals is deficient of a reliable
decision support system factor to get better
solution. A standardization and
centralization of planning procedures which
progresses reliability of the operating theater
and handles the fragile situation of the
waiting lists for surgery was developed using
optimization methods and data mining.
Schizophrenia spectrum disorder was
diagnosis by a decision system created by
Kahn R [11].The system have four-stages:
knowledge acquisition and organization, the
development of computer-aided model, and
evaluation of the overall system’s
performance. The required knowledge is
mined from an professional via interviews.
The knowledge was modeled by algorithms
and transmitted to a reckoning model formed
using covering approach. The performance
valuation is based on comparison of several
diagnoses of eighty one pieces clinical
between DSS and an expert. The outcomes
showed a comparatively small rate of
misclassification (18-34%) also an excellent
performance or behaviour by the DSS with
66-82% accuracy.
Shankar F et. al. [16] proposes a prediction
system for liver disease using intelligent
machine learning. Researches provided
various data to identify the causes for
hepatitis. Here, Decision tree technique was
applied to determine the structural
information of tissues.C4.5 algorithm was
used to generate decision tree that
concentrates on some attributes such like
age, sex, steroids, antivirals, etc for the
diagnosis of the disease. These features
helped in defining the abnormalities of the
sick person with Heart ailment.
4
 Available online at: www.engineeringjournal.info
C5.0 combines several firsthand abilities
such as adaptable misclassification costs. In
C4.5, all faults are classified as alike;
practically some classification faults are more
severe than others. C5.0 allows a distinct
charge to be distinct for each real class pair;
if this choice is used, Then C5.0 makes
classifiers to minimize anticipated
misclassification rates slightly than error
occurrence.
In an application that categorizes individuals
as probable or not probable to "churn," C5.0
efforts to minimize the weighted projected
error rate.
C5.0 is very easy to use. Several possibilities
have been extended and simplified -- to
support sampling and cross-validation,
example -- and C4.5's programs used in
generating rule sets have been fused into a
distinct program.
The Windows form, See5 is very user-friendly
with nice graphic interface.
C4.5 has a standard rule that induces
classification rules in the form of a decision
tree. As an addition of ID3, usual criteria of
selecting splitting attributes in C4.5 are
information gain ratio, in place of using
information gain as that in ID3, information
gain ratio avoids the bias of selecting
attributes with numerous values.
Material and Method
In this research an intensive discussion of a
more accurate intelligent medical prediction
and decision support system model for
diagnosis and prediction of Hepatitis by
using signs and the symptoms of patients as
variable predictors and also making use of
SEE 5.0 decision tree algorithm for
classification. The model was well
implemented using real life sample data of
Hepatitis disease. The data for this research
Figure 2: The use case diagram of the proposed system
LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12
was obtained from University of California
Irvine (UCI) Machine Learning Repository
database.
This work developed a model for an
intelligent medical decision support system
that would aid human medical personnel in
accurate diagnosis and prediction of hepatitis
B using C5.0 algorithm. It will develop a
reliable guide for human physician and
medical personnel decision support system.
The proposed system will produce intelligent
decision support system with ability to
decrease cost and the time of uncertainty
diagnosing hepatitis problems.
The system provides efficient decision process
that involves lots of vagueness and
uncertainty management mostly in a
situation of multiple symptoms. Some of the
benefits of the system includes that; it
minimizes the time constrain for data
classification (It optimizes the system’s time
complexity), the proposed system uses a
command of memory magnitude less than
C4.5 during rule set construction, the C5.0
rule sets formation have noticeably lower
error on unseen cases of various datasets,
C5.0 expresses new processes for creating
rule sets, and the development is substantial.
The use of appropriate method enhances
effectiveness and the efficiency in all
research work. We adopted the object
Oriented Analysis and Design Methodology
in analysis of this intelligent system for
hepatitis diagnoses and treatment. This
methodology features the unified modeling
language (UMLs): Case diagram, Activity
diagram, High Level Model and the
architectural design of the newly introduces
system. The used case diagram can be seen
in Figure 2
5
 Available online at: www.engineeringjournal.info
The use case diagram of the proposed system
can be seen in figures 2 and consist of the
following components:
 C5.0 Algorithm: split the sample
established on the area that gives the
highest information gain. Each subsample
distinguished by the first split is then split
again, usual; Based on a different field and
the process continues reputedly till the
subsample cannot be split anymore.
 Training hepatitis dataset: Training
hepatitis dataset are the theoretical time
series information of various symptoms and
the treatment of hepatitis diseases.
 Intelligent rules for hepatitis diagnoses:
Rule sets proposed in this work hold most
of significant information in C5.0 decision
tree however with less difficult model. The
rules obtain with C5.0 algorithm in union
Figure 3: Activity diagram of the proposed system
The activity diagram of the planned system
as display in figure 3.3 gives the individual
presentation of the varying component of the
proposed system. The raw hepatitis dataset
is gotten from the database. Data mining
model is used to process, integrate
transformed and cleans the extracted raw
hepatitis dataset. Then the C5.0 algorithm
were adapt to thoroughly train the dataset to
formulate rules using the obtained
LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12
with other nodes; will be help to properly
diagnose hepatitis symptoms and
treatment.
 Likelihood of Hepatitis issues: the
intelligent rules developed are used to
predict hepatitis cases and treatment.
 Treatment of Hepatitis: the developed
intelligent rules for hepatitis diagnoses was
adopt to interpret the symptoms from the
client and the identify the treatment and
the type of hepatitis case such as hepatitis
B, C, D or acute Hepatitis.
 Intelligent decision provision system: the
outcomes of our proposed model will
provide intelligent guides for efficient
medical decision for hepatitis treatment
cases.
o The activity diagram of the proposed
system can be seen in Fig. 3.
attributes and symptoms of hepatitis dataset.
The entropy and information again of the
features were compute and the dataset were
been split using the dataset gain ratios. After
the training all the features of trained
dataset, an intelligent rule based were
develop to diagnose hepatitis cases for
efficient treatment and decision quick
making. The architecture of the proposed
system is shown in Figure 4
6
 Available online at: www.engineeringjournal.info

Figure 4: The architecture of the proposed system

The components of architecture of the been used to intelligently diagnose various

proposed system as seen in figure 4 are as
follow:
 Hepatitis Symptoms: the hepatitis
symptoms were used as input to test and
compute the proposed model.
 C4.5 Algorithm: is a machine learning
algorithm were used to develop and
efficient intelligent system to control and
treat hepatitis cases.
 Validated by Domain Expert: the proposed
intelligent system were been integrated
with other component of the system by the
experts.
 Knowledge Base: this contains the
information store in the database that is
cases of hepatitis cases.
 Ruled Based System: developed with C5.0
algorithm are used by the system to
efficiently treat any type hepatitis case
such as hepatitis A, B, C and acute
hepatitis. This rules make prediction for
individual records. Rue sets are derived
from decision trees and, in a way, shows a
simplified or distilled version of the
information seen in the decision tree.
 Display Advice to End User: The outcomes
of intelligent result for hepatitis treatment
were displayed to the client
The high level model is shown in Figure 5

Figure 5: High level model of the proposed system

LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12 7

 Available online at: www.engineeringjournal.info

The High level model of the proposed system mining Model”, “Data warehouse”, and “Data
as depicted in figure 5 comprises of “Data mining techniques”, “Knowledge Discoveries”
gathering from Central Database”, “Data and “Decision Support System”.

Table 1: Training dataset for hepatitis symptoms (Source University of California Irvine
(UCI) Machine Learning Repository database)
Clas AGE SEX STEROIDANTIVIRALS FATIGUE MALAISEANOREXIALIVER_BIG LIVER_FISPLRMEEN_PALPABLSPIEDERS ASCITES VARICES BILIRUBIN ALK_PHOSPHATE SGOT ALBUMIN PROTIME HISTOLOGY Die 54 Male yes no yes yes no no no no no yes no 1.2 85 92 3.1 66 no
Die 50 Male no no yes no no yes no no no no no 0.9 135 42 3.5 ? yes Live 57 Male yes no yes yes no no no no yes yes no 4.6 82 55 3.3 30 no
Die 54 Male no no no no no no no no no no no 1 85 30 4.5 0 no
Die 78 Male yes yes yes no no no no no no no no 0.7 96 32 4 ? yes Live 31 Male yes no yes yes yes no no yes no no no 8 ? 101 2.2 ? no
Die 31 Male no yes no no no no no no no no no 0.7 46 52 4 80 yes Die 48 Male no no yes yes yes no yes no yes no no 2 158 278 3.8 ? no
Die 34 Male yes no no no no no no no no no no 1 ? 200 4 ? yes Die 72 Male no yes yes no no no yes no no no no 1 115 52 3.4 50 no
Die 34 Male no no no no no no no no no no no 0.9 95 28 4 75 yes Live 38 Male yes no no no no no yes no no no no 0.4 243 49 3.8 90 no
Die 25 Male no no yes no no yes yes yes yes yes yes 1.3 181 181 4.5 57 no
Live 51 Male yes no yes no yes no no yes yes no no ? ? ? ? ? yes Die 51 Male no no no no no yes yes no yes no no 0.8 ? 33 4.5 ? no
Die 23 Male no no no no no no no no no no no 1 ? ? ? ? yes Die 38 Male no no no no no no yes no yes no yes 1.6 130 140 3.5 56 no
Die 39 Male no no yes no no no yes no no no no 0.7 ? 48 4.4 ? yes Live 47 Male no no yes yes no no yes no yes yes yes 1 166 30 2.6 31 no
Die 30 Male no no no no no no no no no no no 1 ? 120 3.9 ? yes Die 45 Male no yes no no no no no no no no no 1.3 85 44 4.2 85 no
Die 39 Male yes yes no no no yes yes no no no no 1.3 78 30 4.4 85 yes Die 36 Male yes no yes yes yes yes yes no yes no yes 1.7 295 60 2.7 ? no
Live 54 Male yes no yes yes no no yes yes no yes no 3.9 120 28 3.5 43 no
Die 32 Male no yes yes no no no yes no yes no no 1 59 249 3.7 54 yes Die 51 Male no no yes no no no yes yes yes no yes 1 ? 20 3 63 no
Die 41 Male no yes yes no no no yes no no no no 0.9 81 60 3.9 52 yes Live 49 Male yes no yes yes no no no yes yes no no 1.4 85 70 3.5 35 no
Die 30 Male no no yes no no no yes no no no no 2.2 57 144 4.9 78 yes Live 45 Male no no yes yes yes no no no yes yes no 1.9 ? 114 2.4 ? no
Die 47 Male yes yes no no no no no no no no no ? ? 60 ? ? yes Die 31 Male yes no yes no no no no no no no no 1.2 75 173 4.2 54 no
Live 41 Male no no yes no no no yes yes yes no yes 4.2 65 120 3.4 ? no
Die 38 Male yes no yes yes yes no no no no yes no 2 72 89 2.9 46 yes Live 70 Male yes no yes yes yes no no no no no no 1.7 109 528 2.8 35 no
Die 66 Male no no yes no no no no no no no no 1.2 102 53 4.3 ? yes Die 20 Male yes no no no no no no no no no no 0.9 89 152 4 ? no
Die 40 Male yes no yes no no no yes no no no no 0.6 62 166 4 63 yes Die 36 Male no no no no no no no no no no no 0.6 120 30 4 ? no
Die 38 Male yes no no no no no no no no no no 0.7 53 42 4.1 85 no Live 46 Male no no yes yes yes no no no yes yes yes 7.6 ? 242 3.3 50 no
Die 44 Male no no yes no no no yes no no no no 0.9 126 142 4.3 ? no
Die 38 Male yes yes no no no yes yes no no no no 0.7 70 28 4.2 62 yes
Die 61 Male yes no yes yes no yes yes no yes no no 0.8 75 20 4.1 ? no
Die 22 female no yes yes no no no no no no no no 0.9 48 20 4.2 64 yes Die 53 female yes no yes no no no no yes yes no yes 1.5 81 19 4.1 48 no
Die 27 Male no no yes yes yes yes yes yes yes no no 1.2 133 98 4.1 39 yes Live 43 Male no no yes no no no no yes yes yes no 1.2 100 19 3.1 42 no

The WEKA simulation result for Information Gain Attribute Ratio is shown in Figure 6.

Figure 6: Information gain attribute ratio

C5.0 model works by splitting the sample information gain from WEKA computation.
based on the field that provides the least The information gain value field is displayed
information gain as shown in Figure 7. in Figure 7.

LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12 8

 Available online at: www.engineeringjournal.info

Figure 7: Information gain value field

The C5.0 algorithm slits training data
samples on basis of the highest information
gain value field. The sample subset that is
get from the former split will be split
afterwards. The process will continue until
the sample subject cannot be split and is
usually according to other field. Finally,
examine the lowest level split, those sample
subsets that don’t have remarkable
contribution to the model will be rejected.
The knowledge denoted with decision tree
could be mined and presented using IF
THEN rules. The decision rules for hepatitis
dataset classification is shown in Figure 8.

Figure 8: C5.0 Decision tree for hepatitis classification and control

The intelligence of our proposed system is This application have been run and tested
well defined for proper diagnoses. successfully.
Experiments Carried Out
Rulebase for hepatitis classification.

LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12 9

 Available online at: www.engineeringjournal.info

Table 2: Reveals the comparison of correctly categorized instances of C5.0 and C4.5

Table 2 shows correct classification of instances of C5.0 is 98.05% and C4.5 is

hepatitis datasets. It shows the comparison 94.81%, this implies that C5.0 is more
of correctly classified instances for C4.5, and resourceful than C4.5 algorithm. The results
C5.0 Algorithms. The correctly classified are signified in Figure 9.

Figure 9: Graphical representation of comparison of correctly classified instances for C4.5,

and C5.0.

Table 3: Comparison of time constrain of the process

Table 3 shows the comparison of time seconds and that of C4.5 is 0.05 seconds, this
constrains of process for C4.5, and C5.0 entails that C5.0 algorithm has minimum
machine learning algorithm. The time time complexity than C4.5. The results were
complexity of the process forC5.0 is 0.03 graphically shown in Figure 11.

LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12 10

 Available online at: www.engineeringjournal.info

Figure 10: Graphical representation of the comparison of Time constrain of the process for
C5.0, and C4.5.

Table 4: Comparison of mean absolute error for C5.0 and C4.5 algorithm classification
process.

Table 4 shows the comparison of Mean 0.09%.This means C5.0 algorithm has the
Absolute Error for C5.0 and C4.5decision tree minimum mean absolute error of
categorization process. The Mean Absolute classification than C4.5 algorithms. The
Error for C5.0 is 0.04while that ofC4.5 is results were represented graphically in
Figure 11.

Figure 11: Comparison of mean absolute error for classification

Discussion of Results
In this work WEKA (simulator software for
data mining) was used to simulate C5.0 to
develop a decision support system for
hepatitis diagnoses and treatment. The
outcomes of the simulation were then used to
create an intelligent system (decision
support) for treatment of hepatitis cases.
100% of the training data was used with C5.0
algorithm to design rules and pattern to
control various complicated cases of hepatitis
problems.
The intelligent system was developed to
interpret the symptoms of the hepatitis
diagnoses and treatment. Using the
simulator (WEKA) we obtain the attribute of
a specific tree node using the C5.0
information pain, split information and gain
ration for C5.0. The simulation result shows
that the Mean Absolute Error for C5.0 is 0.04
while that of C4.5 is 0.09% (see Table 4). This
implies that C5.0 algorithm has the
minimum mean

LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12 11

 Available online at: www.engineeringjournal.info

absolute error of classification than C4.5 Conclusion

algorithms. The time complexity of the
This work have successfully designed an
process for C5.0 is 0.03 seconds while that of
intelligent Medical decision support system
C4.5 is 0.05 seconds. This implies that C5.0
using C5.0 machine learning algorithm that
algorithm has minimum time complexity
will help to improve treatment of various
than C4.5 (see Table 3). The correctly
cases of Hepatitis challenges. The developed
classified instances of C5.0 is 98.05% while
intelligent system improved the uncertainty
C4.5 is 94.81%, this implies that C5.0 is more
in the estimation of the treatment of various
efficient than C4.5 algorithm (see Table 2)
cases of hepatitis. It provides a reliable guide
for hepatitis A and B treatment for
physicians and medical personnel.

References
1 Ahmed (2014) Neutrosophic and fuzzy C/ means Evans and V. Petal (Eds.), Cognitive Science
gathering for enhancing CT liver appearance. and Medicine: Biomedical Modeling, 11-255.
IEEE Transactions of Information Technology MIT Press, Cambridge, Massachusetts.
in Biomedicine, 11(4):101-113
10 Gt Venira, Emeksiz (2010) presented an expert
2 Amin SU, Agarwal K, Beg R (2013) Data system for differential diagnosis’, American
mining in clinical decision support system’. Journal of Computer Sciences and
International Journal of Advanced Research in Mathematics, 129(2):73-85
Computer Engineering & Technology
11 Kahn R, Perkins D, Lieberman J (2013)
(IJARCET), 2(1):56-67.
Predictors of treatment response in patients
3 Anil, Anithia (2015) "A soft computing with first-episode prostate cancer disorder. The
paradigm for a medical data mining tool to British Journal of Gynecology, 185(1):18-24.
predict risk of coronary heart events", The
12 Ming, R, Pale T (2010) Applied neutrosophy to
British Journal of Medical Science, 121(1):21-
image processing. International Journal of
32.
Advanced Research in Computer Engineering &
4 Bekir (2011) Comparison between Back Technology (IJARCET), 1(3):106-119.
propagation and Naive Baye Classifiers to
13 Nicolae V, Lath D (2012) an approach to
diagnose Hepatitis disease’, International
integrate computational intelligence with
Journal Computer Engineering and applied
clinical investigation methods and knowledge
Physics. 60(2):97-118
for glaucoma diagnosis’.American Journal of
5 Brause (2008) Physicians Roll in Clinical Applied Sciences, 179(2), 98-112
Decision Support Systems: State of Art.
14 Oguntimilehin A, Adetunmbi AO, Abiola OB
Rockville: AHRQ Publication 10(4):50-70.
(2015) “A Machine Learning Approach to
6 Chrysostomos (2008) A new hybrid modeling Clinical Diagnosis of Typhoic Fever”
methodology for complex decision making International Jounal of Computer and
procedures’. American Journal of Computer Information Technology, 2(4).
Sciences and Mathematics, 89(3):68-82
15 Sonawane, Vadicherla (2013) Program based
7 Djain and Kimbi (2011) proposed a medical Medical Decision Support System (MDSS).
diagnostic support system for the management IEEE Transactions of Information Technology
of hypertension’International Journal of in Biomedicine, 9(7):205-215
Advanced Research in Computer Engineering &
16 Shankar F, Gomes C, Brito A (2014) A
Technology (IJARCET), 1(5):126-147.
prediction system for liver disease using
8 Dragulescuand Albu (2007) Logical Inference intelligent machine learning Automation
System For detection of Hepatitis virus’Jounal Science and Engineering, IEEE Transactions on
of Computer and Information Technology, 1(3). Robotics & Control Systems, 99-112.
9 Evan DA, Gadd CS (2005) Managing coherence
and context in medical problem solving. In D.

LN Onyejegbu et. al.|June 2018| Vol.03 | Issue 06 |01-12 12