1114 IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 14, NO.
4, JULY 2010
Intelligible Support Vector Machines for Diagnosis
of Diabetes Mellitus
Nahla H. Barakat, Andrew P. Bradley, Senior Member, IEEE, and Mohamed Nabil H. Barakat
Abstract—Diabetes mellitus is a chronic disease and a major
public health challenge worldwide. According to the International
Diabetes Federation, there are currently 246 million diabetic peo-
ple worldwide, and this number is expected to rise to 380 million
by 2025. Furthermore, 3.8 million deaths are attributable to di-
abetes complications each year. It has been shown that 80% of
type 2 diabetes complications can be prevented or delayed by early
identification of people at risk. In this context, several data min-
ing and machine learning methods have been used for the diag-
nosis, prognosis, and management of diabetes. In this paper, we
propose utilizing support vector machines (SVMs) for the diag-
nosis of diabetes. In particular, we use an additional explanation
module, which turns the “black box” model of an SVM into an
intelligible representation of the SVM’s diagnostic (classification)
decision. Results on a real-life diabetes dataset show that intelligi-
ble SVMs provide a promising tool for the prediction of diabetes,
where a comprehensible ruleset have been generated, with predic-
tion accuracy of 94%, sensitivity of 93%, and specificity of 94%.
Furthermore, the extracted rules are medically sound and agree
with the outcome of relevant medical studies.
Index Terms—Data mining, diabetes, machine learning, medical
diagnosis.
I. INTRODUCTION
IABETES is a disease primarily associated with an in-
D crease in the level of blood glucose (hyperglycaemia) [1].
One reason for hyperglycaemia is insulin deficiency, where beta
cells in the pancreas fail to produce enough insulin. This is
known as type 1 diabetes. The other and most common type of
diabetes is recognized as type 2, where the body cannot effec-
tively use the insulin produced [2].
Chronic hyperglycaemia in people with diabetes increases
the risk of microvascular damage, which leads to retinopathy,
nephropathy, and neuropathy. Therefore, diabetes is the leading
cause of blindness and visual impairment in adults in devel-
oped countries [2] and is responsible for over one million lower
limb amputations each year. Diabetic people are also exposed
to an elevated risk of macrovascular complications, where they
are two to four times more likely to get cardiovascular disease
(CVD) than people without diabetes. Due to these complica-
Manuscript received June 6, 2009; revised October 20, 2009; accepted
December 16, 2009. Date of publication January 12, 2010; date of current
version July 9, 2010.
N. H. Barakat is with the Department of Applied Information Technol-
ogy, German University of Technology in Oman, Muscat 130, Oman (e-mail:
n.barakat@uq.edu.au).
A. P. Bradley is with the School of Information Technology and Electrical
Engineering, University of Queensland, Brisbane, QLD 4072, Australia (e-mail:
bradley@itee.uq.edu.au).
M. N. H. Barakat is with the Department of the Noncommunicable Diseases
Surveillance and Control, the Ministry of Health, Muscat 113, Oman (e-mail:
mnbarkat@yahoo.co.uk).
Digital Object Identifier 10.1109/TITB.2009.2039485
1089-7771/$26.00 © 2010 IEEE
tions, diabetes is found to be the fourth leading cause of global
death by disease.
The prevalence of type 2 diabetes is increasing at a fast pace
due to obesity, in particular, central obesity, physical inactivity,
and unhealthy dietary habits [3]. Early detection of diabetes
would be of great value given the fact that at least 50% and
80% in some countries, of all people with diabetes are unaware
of their condition and will remain unaware until complications
appear [2], [4].
Recent studies have shown that 80% of type 2 diabetes com-
plications can be prevented or delayed by early identification and
intervention in people at risk [2], [4], for example, by chang-
ing their lifestyle [3] and/or by therapeutic methods. Intelligent
data analysis, such as data mining and machine learning tech-
niques are, therefore, valuable for identifying those people. In
this context, several data mining and machine learning methods
have been proposed for the diagnosis and management of dia-
betes [5]–[9]. It has also been shown that employing computer-
aided diagnostic systems (CAD) as a “second opinion” has lead
to improved diagnostic decisions [10], and support vector ma-
chines (SVMs) have shown remarkable success in this area [11].
In this study, we propose a novel hybrid model for medical di-
agnosis, integrating three different data mining/machine learn-
ing techniques. In particular, an unsupervised and supervised
learning algorithm are used for sampling and model building,
respectively, which are then followed by a rule-based explana-
tion component. Furthermore, we validate our model using a
real-life dataset for the prediction of type 2 diabetes. As we will
demonstrate, our technique is able to predict diabetes with high
accuracy, sensitivity, and specificity, which outperforms other
techniques [8], [9] working on relevant problems. We also show
that the diagnostic criteria learned by our model are valid from
a medical stand point and that our results are supported by other
medical studies [12].
The paper is organized as follows. A brief background
for SVMs and rule extraction from SVMs is provided in
Section II. The experimental methodology is presented in
Section III, followed by results and discussion in Section IV.
Rule interpretation and validation is demonstrated in Section V
and some conclusions are drawn in Section VI.
II. BACKGROUND
A. Support Vector Machines
SVMs operate by finding a linear hyperplane that separates
the positive and negative examples with a maximum interclass
distance or margin d. In the case of unequal misclassification
costs, a cost factor J (C+ /C− ) is introduced by which training
BARAKAT et al.: INTELLIGIBLE SUPPORT VECTOR MACHINES FOR DIAGNOSIS OF DIABETES MELLITUS
errors on positive examples outweigh errors on negative exam-
ples [13]. Therefore, the optimization problem becomes
1

prune rules with performance below a user-defined threshold.
minimize
w
2 + C+ ξi + C− ξj
2 i:y =1i j :y =−1 j
Subject to yk (wxk + b) ≥ 1 − ξk , ξk ≥ 0. (1)
where yi is the class label, w is normal to the hyper-plane,
|b|/
w
is the perpendicular distance from the hyper-plane to
the origin,
w
is the Euclidean norm of w, C is a regularization
parameter, which defines the tradeoff between the training error
and the margin d, and ξ i is a slack variable to allow errors in
classification [13].
To handle nonlinearly separable data kernel functions are
used, including kernel functions and a Lagrange multiplier αi ,
the dual optimization problem becomes

learned by the SVM are then extracted from the synthetic dataset
1

l l
maximize w(α) = αi − αi yi αj yj K (xi .xj )
i=1
2 i=1,j =1


l
C ≥ αi ≥ 0 ∀i , αi yi = 0. (2)
i=1
Solving for α, training examples with a nonzero α are called
support vectors (SVs) and the hyper-plane is completely defined
by the SVs alone.
There is, however, a significant drawback to SVMs, in that
they have an inability to provide a comprehensible justification
for the classification decisions they make. That is, they are black
box models. In medical diagnosis, it has been shown that the
explanation of a classification decision is a crucial requirement
for the acceptance of black box models by end users [14]–[16].
Therefore, techniques for rule extraction have been introduced
[17] to enable SVMs to be more intelligible.
B. Rule Extraction From SVMs
Direct rule learners extract rules that describe a pattern or
relationships between input features and output class labels di-
rectly from the data. In the case of black box models, these
relationships are not comprehensible to end users. Therefore,
the task of rule extraction is to devise rules from the model,
rather than directly from the data. In doing so, an explanation
of the knowledge learned by the black box model (from the
data) and embedded in the structure of the model (SVs in the
case of SVMs) is revealed and provided to the end users in a
comprehensible form.
In this paper, we utilize two different techniques for rule
extraction as the last module of our proposed approach, in par-
ticular SQRex-SVM [18] and the eclectic [19] methods are used
to turn the SVM black box into a more intelligible model. In
the following subsections, a brief description of these methods
is introduced.
1) SQRex-SVM: A Sequential Covering Approach for Rule
Extraction: SQRex-SVM [18] extracts rules directly from a
subset of the SVM SVs [see (2)], using a modified sequential
covering algorithm and based on an ordered search of the most
discriminatory features. These features are then ranked and uti-
1115
lized to form an initial ruleset for the positive class, which is
then refined and pruned. A prepruning strategy is adopted to
In the postpruning step, the algorithm utilizes the area under the
receiver operating characteristic (ROC) curve (AUC) to con-
trol the tradeoff between the classifier (ruleset) performance, in
terms of both the true positive (TP) and false positive (FP) rates
and AUC, and comprehensibility as measured by the number of
rules. Rules that do not result in a statistically significant (p <
0.05) increase in the ruleset AUC are pruned.
2) Eclectic Rule Extraction: In this approach [19], a labeled
dataset is used to train an SVM to obtain a model (classifier)
with acceptable accuracy, precision, and recall. Next, a synthetic
dataset composed of the training examples that became SVs is
constructed with the target class for these examples replaced by
the class predicted by the SVM. Rules representing the concepts
using the C5 decision tree learner [20].
III. EXPERIMENTAL METHODOLOGY
A. Dataset
The dataset used in this paper has previously been used in
[21], which investigated the prevalence of diabetes in Oman.
Another study on this data [22] used data mining methods to
learn rules for the diagnosis of diabetes. However, in this paper,
we will be employing SVMs for the first time for the diagnosis
of diabetes and justifying the SVM’s classification decisions by
the extracted rules. A detailed description of the dataset can be
found in [21]. However, for convenience, a brief description is
provided next.
Data from 4682 subjects of age 20 years and above was col-
lected using a questionnaire regarding demographic data, his-
tory, and anthropometric measures. Furthermore, blood pressure
and blood samples were analyzed to measure fasting venous and
2-h postglucose load [oral glucose tolerance test (OGTT)] for
the participants of the survey in healthcare centers. Other at-
tributes were also collected, but omitted here as they are not
relevant to this study.
The data collected about each subject include age in years
(min 20 and max 80), sex (male/female), family history of di-
abetes (yes/no), body mass index (BMI) (min 14 and max 47),
waist circumference in centimeter (waist) (min 45 and max
120), hip circumference (min 66 and max 125 cm), systolic
blood pressure (min 90 and max 220), diastolic blood pressure
(BPDIAS) (min 50 and max 120), cholesterol (min 1.9 and
max 8.5 mmol/L), fasting blood sugar (FBS) (min 55 and max
320 mg/dL), 2 h postglucose load (OGTT) min 38 and max
570 mg/dL).
The diabetes detection method used in this dataset is the
OGTT, according to the 1985 World Health Organization
(WHO) criteria, where the subject is considered diabetic if their
OGTT is equal to or greater than 200 mg/dL. Diabetic subjects
who were taking oral hypoglycemic tablets or insulin (i.e., which
were diagnosed as diabetic prior to the survey) were excluded
from the study. In addition, all subjects with missing values in
1116 IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 14, NO. 4, JULY 2010
TABLE I
DATASET
any of the 11 features were excluded. As a result, there are a
total of 3014 subjects in the dataset.
B. Preprocessing and Sampling
The prevalence of diabetes in the dataset was about 9%, which
means that the class priors of the dataset are highly skewed
toward nondiabetic subjects (negative class).
We have selected a representative sample for the majority
class (nondiabetic subjects) using subsampling and a K-means
clustering algorithm. Subsampling was used to overcome the
problem of extremely skewed class distribution [23], where it
has been shown that class imbalance may not allow the learn-
ing algorithms to learn good models [24], [25]. Furthermore,
decreasing the number of negative examples in the training set
makes the learned model more sensitive to false positive classi-
fications.
The clustering algorithm was employed for subsampling to
ensure the selection of a representative training set, especially
for the negative (majority) class. This was done because select-
ing training examples at random may result in important patterns
in the original dataset being excluded from training set, hence,
impeding the quality of the learned model.
The K-means [23] clustering algorithm was used as follows.
1) Five clusters were devised from the original dataset (as
only 20% of negative examples were required).
2) Within each cluster, samples (subjects) were sorted based
on their Euclidean distance from the cluster center.
3) Samples at the maximum distance from the cluster center
were excluded, as they are likely to be outliers.
4) From each of the sorted clusters, every third sample is
included in the final dataset, to select representative ex-
amples from each cluster, at different Euclidean distances
from the center.
5) All positive samples were included in the final dataset.
6) The selected subjects in the final dataset were then ran-
domly divided into two disjoint subsets: training and test-
ing. Details of the training and testing sets are shown in
Table I.
C. SVM Training
The first ten features (risk factors) were used to train the SVM
to predict the diagnosis [diabetic (class label 1), nondiabetic
(class label-1)]. Specifically, a subject is considered diabetic if
his/her OGTT is ≥200 mg/dL, and nondiabetic if the OGTT is
<200 mg/dL.
The SVMlight [26] implementation was used in all experi-
ments. The SVM training parameters were selected using leave-
one-out cross validation on the training set. A radial basis func-
TABLE II
RULE PERFORMANCE COMPARED TO THE SVM AND DIRECT RULE LEARNERS
AT EQUAL MISCLASSIFICATION COSTS
tion (RBF) kernel, with γ (gamma) of 0.0005 and C (regular-
ization parameter) of 5 were chosen, as they gave the best recall
with the lowest error rate on the training set [see (1)].
The training procedure described in [19] was used as follows.
1) A number of SVM models were generated by varying the
misclassification cost factor J (1), starting with a small
value and increasing J until no change in TP or FP rates
was observed.
2) Each of the generated models were then used to classify
the independent test set and accuracy, TP and FP rates
calculated.
3) Rules were then extracted from each of the models
using SQRex-SVM and eclectic methods described in
Section II.
4) Each of the extracted rulesets were then used to classify
the same independent test set and again accuracy, TP and
FP rates, as well as fidelity were computed.
5) ROC curves were then plotted for both the SVM and rule-
sets and the AUC was computed using trapezoidal inte-
gration. Standard errors for the AUC were estimated via
the standard error of the Wilcoxon statistic [19].
IV. RESULTS AND DISCUSSION
Table II shows accuracy results of the SVM, the rule extrac-
tion methods and three direct rule learners [20], [27], [28] at
equal misclassification costs, as well as rule fidelity (the extent
to which the prediction behavior of a ruleset mimics that of the
SVM from which the rules were extracted) and comprehensi-
bility (as measures by number of rules/antecedents).
Comparing the accuracy of the extracted rules to the SVM,
it can be seen that the rules extracted by the SQRex-SVM
and eclectic approaches have better accuracy than that of the
SVM. The best accuracy is obtained by direct rule learning us-
ing C5 [20]. However, the difference in accuracy between C5
rules and those of SQRex-SVM is not statistically significant
(p > 0.05). Furthermore, the SQRex-SVM ruleset has the best
comprehensibility.
Table III shows the TP and FP rates for the rulesets extracted
from the SVM and those of the SVM and direct rule learners.
From these results, it can be seen that the eclectic approach
achieved the best TP rate and AUC, followed by SQRex-SVM.
To determine if the difference in AUC between the SQRex-
SVM and the eclectic approaches is statistically significant, a
large sample z-test was performed. Results indicate that the null
BARAKAT et al.: INTELLIGIBLE SUPPORT VECTOR MACHINES FOR DIAGNOSIS OF DIABETES MELLITUS
TABLE III
RULE TP, FP RATES, AND AUC COMPARED TO THE SVM AND DIRECT RULE
LEARNERS AT EQUAL MISCLASSIFICATION COSTS
Fig. 1. ROC curves for the eclectic rules and the SVM.
hypothesis shows that there is no difference in measured AUC
between the two approaches cannot be rejected (p > 0.05).
Furthermore, the differences in AUC between the SVM and
those of the eclectic and the SQRex-SVM approaches are not
statistically significant (p > 0.05). It should be noted that the
same training and testing sets described in Table I have been
used to train the SVM and direct rule learners, and the same test
set is used to assess the quality of the SVM model, the extracted
rules and the rules obtained by direct rule learners.
The ROC curves for the performance of the SVM and the
extracted rules on the test set at different misclassification costs
are shown in Figs. 1 and 2, respectively.
Comparing the ROC curves for the SVM and the extracted
rulesets, it can be seen that both curves follow the same pattern
with increasing misclassification cost. It can also be seen that
the ROC curves for both the eclectic and SQRex-SVM methods
are almost identical to the SVM ROC curves, which is also
reflected by their AUC. It should be noted here that the ROC
curve had to be manually connected to the point (1,1) for the
SVM and extracted rulesets (as no value of J ensured the SVM
always predicted positive).
The AUC for these ROC curves and the associated standard
errors, as well as those direct rule learners are shown in Table IV.
From this table, it can be shown that C5 has slightly higher
AUC than rules extracted from SVM. However, this difference
1117
Fig. 2. ROC curves for SQRex-SVM rules and the SVM.
TABLE IV
RULES AUC AT DIFFERENT MISCLASSIFICATION COSTS
in AUC is not statistically significant (p > 0.05). Furthermore,
the rulesets extracted from SVMs have smaller number of rules,
and therefore, improved comprehensibility. In addition, the dif-
ference between the AUC of rules extracted from the SVM by
the two approaches, as well as the difference of AUC between
each approach and the original SVM are not statistically signif-
icant (p > 0.05).
We have also compared our results with those of similar
studies, e.g., [8], [9], where classification and regression tree
(CART) decision tree [27] has been used for the prediction of,
or people at risk of, diabetes; it can be seen that our approach
has obtained improved results. The best results obtained in [9]
were 88%, 75%, and 85%, while best results obtained in [8]
were 94.5%, 38%, and 73.6% for sensitivity, specificity, and
AUC, respectively.
V. DIAGNOSTIC RULES VALIDATION AND INTERPRETABILITY
After confirming the quality of the rules extracted by the two
methods and showing that they are providing a succinct explana-
tion to the concepts learned by SVMs, we now look at a major
complementary measure of rule quality, rule interpretability,
and whether these rules make sense to domain experts.
In principle, the rules extracted by the two methods have
shown the correct and valid risk factors (features), which are
used for the diagnosis and prediction of diabetes in their an-
tecedent (namely, FBS, BPDIAS, and waist circumference).
1118 IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 14, NO. 4, JULY 2010
However, there are slight variations in the cutoff values between
the methods. One possible reason for this variation is the differ-
ent induction bias for each rule extraction method. Specifically,
in the eclectic approach, a decision tree learner is used, which
in turn uses information gain to evaluate both the best feature
to split on and the preferred threshold value. Information gain
is not the measure used in SQRex-SVM, which uses the TP and
FP rates. Therefore, while both methods have selected the same
features, they have set their preferred threshold values slightly
differently. A more detailed description of the extracted rules
and their validation is provided in the following subsections.
A. SQRex-SVM Rules
It was shown in Section IV that SQRex-SVM extracts the best
rules in terms of rule quality from the machine learning point
of view, i.e., simplest and of best accuracy. It will therefore be
interesting to see if these rules have same quality when evaluated
from the medical expert’s perspective.
The following are the rules extracted by SQRex-SVM at equal
misclassification cost.
1) IF FBS >106.2
Then diabetic.
2) IF waist circumference ≥ 91
and BPDIAS ≥ 90
Then diabetic.
Default rule, nondiabetic.
If we consider the first rule, this rule is valid from the medical
perspective, as the diagnosis of diabetes can be solely done by
the FBS [1]. Considering the cutoff value of 106.2, this value is
supported by a previous study [29], which used two datasets (one
of them being the data used in this study) and concluded that
this value (106.2 mg/dL), even though it is lower than the cutoff
value defined by the American Diabetes Association (ADA) and
WHO for the diagnosis of diabetes [1], is the value that gave
best AUC, sensitivity, and specificity for the Omani population,
hence, it is the value that best matches an OGTT ≥ 200 mg/dL.
Al-Lawati and Barakat [29] have pointed out that if the ADA
cutoff is applied, it would lead to underestimation of diabetes
by (18%) in the Omani population. Similar studies [30]–[32]
have also shown that the ADA criteria would also underestimate
diabetes in other populations. This rule can also be used to
discover undiagnosed diabetic subjects, missed by ADA and
WHO criteria.
Considering the second rule, recent studies have shown an
association between raised blood pressure (hypertension) and
central obesity, which is mainly measured by waist circum-
ference and the risk of developing diabetes and/or metabolic
syndrome [33]. A risk score for developing diabetes has been
proposed in [12], where hypertension contributed three points
and waist circumference contributed two points out of ten-point
scale for all diabetes risks factors. Furthermore, subjects with
metabolic syndrome also face an elevated risk of the develop-
ing type 2 diabetes and CVDs [34]. The International Diabetes
Federation (IDF) defines metabolic syndrome as the presence
of central obesity, plus any two of the following factors.
Fig. 3. ROC curve for waist circumference as predictor for hyperglycemia.
1) Raised triglyceride level.
2) Reduced high-density lipoprotein cholesterol.
3) Raised blood pressure (systolic blood pressure ≥ 130 or
BPDIAS ≥ 85 mm·Hg).
4) Raised fasting plasma glucose (FPG ≥ 100 mg/dL, or
previously diagnosed as type 2 diabetes).
As mentioned earlier, central obesity is measured by waist
circumference, but it is also ethnicity specific [33].
Comparing the IDF cutoff points for waist circumference with
the one obtained by the second rule, it can be seen that the cutoff
point we obtained for waist circumference is valid, as it is close
to the values defined by IDF [33] for different ethnic groups.
It should also be noted that there is no cutoff value specifically
defined for the Middle East (Arab) population [33]; and there is
a recommendation to use the European data until more specific
data becomes available.
To further validate the second rule, we have plotted two ROC
curves for the waist circumference as a predictor of the two
most common risk factors for metabolic syndrome namely, hy-
perglycemia and hypertension as shown in Figs. 3 and 4. The
waist circumference thresholds (cutoff values) used to plot these
ROC curves were <75, ≥75 and <80, ≥80 and <85, ≥85 and
<91, and ≥91 cm.
From these figures, it can be seen that waist circumference of
91 is a better predictor of each of these risk factors as compared
to lower waist circumference. This is more evident in males than
females (p < 0.05). We have also investigated the distribution of
diabetic subjects (as defined by the WHO criteria) over different
groups of waist circumferences. As Fig. 5 shows the distribution
confirms our results, where the highest percentage of diabetic
subjects have a waist circumference ≥91 cm, followed by the
range of waist circumference ≥85 cm.
Based on the discussion earlier, it can be concluded that the
second rule can be used for opportunistic screening to identify
people at risk [12] (predicting diabetes). If indicated, further
investigation should be carried out and perhaps starting an early
intervention program to control the development of diabetes [3].
BARAKAT et al.: INTELLIGIBLE SUPPORT VECTOR MACHINES FOR DIAGNOSIS OF DIABETES MELLITUS
Fig. 4. ROC curve for waist circumference as predictor for hypertension.
Fig. 5. Distribution of diabetic subjects over waist circumference groups.
Figs. 3–5 also validate our rule, which suggest that the cutoff
value of 91 cm can be used as a guiding or starting point for
the definition of central obesity in both males and females in
the Arabian Gulf area population. It can also be expected that
a cutoff value, which could be between 85 and 91 cm for both
males and females is more suitable for defining central obesity
as compared to the cutoff value of Europeans suggested by the
IDF [33]. However, more retrospective studies are needed to
decide the best cutoff value.
It should be noted that the cutoff values for FBS as well as
the waist circumference in our rules, which mainly diagnose
diabetes were the same for both males and females, which is
also the case in the WHO and IDF. However, we have plotted
separate ROC curves for males and females (see Figs. 3 and
4), for predicting different risk of metabolic syndrome, where
IDF standards distinguish between males and females regarding
waist circumference cutoff values. Unlike the IDF standards,
our results suggest that the waist circumference cutoff value for
predicting the risk of metabolic syndrome is the same for both
1119
males and females. However, there are other studies, which also
did not distinguish between males and females as well [35].
It is also worthwhile mentioning here that the average waist
circumference of males and females is the same in the survey
data, which explains the same cutoff value for both genders.
B. Eclectic Method Rules
The following are the rules extracted by the eclectic approach.
1) IF FBS > 124.2
Then diabetic.
2) IF FBS > 90
and FBS ≤ 124.2
and waist circumference > 84
and BPDIAS > 70
Then diabetic.
It should be noted that there were an additional three rules for
nondiabetics, which were removed as they were basically the
complement of the aforementioned rules.
Considering these rules it can be seen that the same risk
factors, namely, FBS, waist circumference, and BPDIAS also
appearing as rule antecedents. Although there is a difference in
the cutoff values due to the reasons mentioned earlier, the rules
(especially the first rule, which agrees with the WHO cutoff)
are still valid from a medical perspective. However, they do
not have such good results as the SQRex-SVM method. This
finding is again supported by [29] for this specific population.
The second rule particularly is not consistent with generally
accepted medical knowledge for the diagnosis of diabetes, as the
cutoff value for the BPDIAS is lower than the one defined for
hypertension. Therefore, this could be a redundant antecedent.
However, this rule can be again used for screening of people at
risk or undiagnosed subjects.
VI. CONCLUSION AND FUTURE PERSPECTIVES
In this paper, we have developed a hybrid system for medical
diagnosis. In particular, we have employed SVMs for the diag-
nosis and prediction of diabetes, where an additional rule-based
explanation component is utilized to provide comprehensibility.
The SVM and the rules extracted from it, are intended to work as
a second opinion for diagnosis of and as a tool to predict diabetes
through identifying people at high risk. According to domain
experts, the significance of our approach lies in its simplicity,
comprehensibility, and validity. They find the rules produced
by the system really helpful, where simple measurements in
the outpatient could be used for opportunistic screening. This
would offer an enhanced opportunity for timely and appropriate
intervention to take place, which would reduce or control the in-
cidence of diabetes or its expensive complications. Furthermore,
the rules can help in the detection of undiagnosed subjects.
Results show that our model is of high quality in terms of
diagnostic and prediction accuracy, and outperforms other tech-
niques working on similar problems.
One of the potential future extensions of this work is
to conduct a prospective study to further refine the predic-
tive results obtained by the proposed rules. Specifically, this
could be achieved by following up the subjects with a waist
1120 IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 14, NO. 4, JULY 2010

circumference ≥91 cm and/or BPDIAS ≥90 for a 3 to 5 year [22] M. N. Barakat, N. Barakat, J. Diederich, and J. Al Lawati, “Diagnosis
period to see if, and when, they develop diabetes. Based on of diabetes mellitus: A data mining approach,” Int. J. Diabetes Metab.,
vol. 13, no. 1, p. 42, 2005.
the outcomes of such a study, a more sophisticated risk score [23] P. N. Tan, M. Steinbach, and V. Kumar, Introduction to Data Mining.
could be developed, which could significantly decrease health- New York: Addison-Wesley, 2005.
care costs via early prediction and diagnosis of diabetes. [24] N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “SMOTE:
Synthetic minority over-sampling technique,” J. Artif. Intell. Res., vol. 16,
pp. 321–357, 2002.
REFERENCES [25] N. Japkowicz, “The class imbalance problem: Significance and strategies,”
in Proc. Int. Conf. Artif. Intell. ({IC}-{AI} 2000), pp. 111–117.
[1] WHO/IDF 2006 (2007, Jan.). Definition and diagnosis of dia-
[26] T. Joachims, Learning to Classify Text Using Support Vector Machines.
betes mellitus and intermediate hyperglycemia, World Health Or-
Norwell, MA: Kluwer, 2002.
ganisation [Online]. Available: http://www.who.int/diabetes/publications/
[27] L. Breiman, J. Friedman, R. Olshen, and C. Stone, Classification and
Definition%20and%20diagnosis%20of%20diabetes_new.pdf
Regression Trees. Monterrey, CA: Wadsworth and Brooks, 1984.
[2] International Diabetes Federation, Diabetes Atlas, 3rd ed. Brussels, Bel-
[28] I. Witten and E. Frank, Data Mining: Practical Machine Learning Tools
gium: International Diabetes Federation, 2007.
and Techniques, 2nd ed. San Francisco, CA: Morgan Kaufmann, 2005.
[3] M. Uusitupa, “Lifestyle matter in prevention of type 2 diabetes,” Diabetes
[29] J. A. Al-Lawati and M. N. Barakat, “Fasting cut-points in determining
Care, vol. 25, no. 9, pp. 1650–1651, 2002.
prevalence of diabetes in an Arab population of the middle east,” Diabetes
[4] M. Franciosi, G. D. Berardis, M. C. E. Rossi, and M. Sacco, “Use of
Res. Clin. Prac., vol. 75, no. 2, pp. 241–245, 2007.
the diabetes risk score for opportunistic screening and impaired glucose
[30] M. I. Harris, R. C. Eastman, C. C. Cowie, K. M. Flegal, and M. S.
tolerance,” Diabetes Care, vol. 28, no. 5, pp. 1187–1193, 2005.
Eberhardt, “Comparison of diabetes diagnostic categories in the U.S. pop-
[5] Y. Huang, P. McCullagh, N. Black, and R. Harper, “Feature selection
ulation according to the 1997 American Diabetes Association and 1980–
and classification model construction on type 2 diabetic patient’s data,”
1985 World Health Organization diagnostic criteria,” Diabetes Care,
in Lecture Notes Artificial Intelligence, vol. 3275, P. Perner, Ed. Berlin,
vol. 20, no. 12, pp. 1859–1862, 1997.
Germany: Springer-Verlag, 2004, pp. 153–162, ICDM 2004.
[31] P. W. Wahl, P. J. Savage, B. M. Psaty, T. J. Orchard, J. A. Robbins, and
[6] R. Bellazzi, C. Larizza, P. Magni, S. Montani, and M. Stefanelli, “Intelli-
R. P. Tracy, “Diabetes in older adults: Comparison of 1997 American
gent analysis of clinical time series: An application in the diabetes mellitus
Diabetes Association classification of diabetes mellitus with 1985 WHO
domain,” Artif. Intell. Med., vol. 20, pp. 37–57, 2000.
classification,” Lancet, vol. 352, no. 9133, pp. 1012–1015, 1998.
[7] R. Bellazzi, “Telemedicine and diabetes management: Current challenges
[32] J. E. Shaw, M. D. Courten, E. J. Boyko, and P. Z. Zimmet, “Impact of new
and future research directions,” J. Diabetes Sci. Technol., vol. 2, no. 1,
diagnostic criteria for diabetes on different populations,” Diabetes Care,
pp. 98–104, 2008.
vol. 22, no. 5, pp. 762–766, 1999.
[8] R. Goel, A. Misra, D. Kondal, R. M. Pandey, N. K. Vikram, J. S. Wasir,
[33] K. G. M. M. Alberti, P. Zimmet, and J. Shaw, “Metabolic syndrome—A
V. Dhingra, and K. Luthra, “Identification of insulin resistance in
new worldwide definition. A Consensus Statement from the International
Asian Indian adolescents: Classification and regression tree (CART) and
Diabetes Federation,” Diabet. Med., vol. 23, pp. 469–480, 2006.
logisticregression-based classification rules,” Clin. Endocrinol., vol. 70,
[34] S. Grundy, “Obesity, metabolic syndrome, and cardiovascular disease,”
pp. 717–724, 2009.
J. Clin. Endocrinol. Metab., vol. 89, no. 6, pp. 2595–2600, 2004.
[9] K. E. Heikes, B. Arondekar, D. M. Eddy, and L. Schlessinger, “Diabetes
[35] H. Wahrenberg, K. Hertel, B.-M. Leijonhufvud, L.-G. Persson, E. Toft, and
risk calculator, a simple tool for detecting undiagnosed diabetes and pre-
P. Arner, (2005). Use of waist circumference to predict insulin resistance:
diabetes,” Diabetes Care, vol. 31, no. 5, pp. 1040–1045, 2008.
Retrospective study, Brit. Med. J., viewed Jan. 2007. [Online]. Available:
[10] N. Lavrac, E. Keravnou, and B. Zupan, “Intelligent data analysis in
http://www.bmj.com/cgi/rapidpdf/bmj.38429.473310.AEv1
medicine,” in Encyclopedia of Computer Science and Technology, vol. 42,
A. Kent et al., Ed. New York: Marcel Dekker, 2000, pp. 113–157.
[11] W. Kong, L. Tham, K. Y. Wong, and P. Tan, “Support vector machine
approach for cancer detection using amplified fragment length polymor-
phism (AFLP) method,” presented at the the 2nd Asia-Pac. Bioinformatics
Conf. (APBC 2004), Dunedin, New Zealand. Nahla H. Barakat received the Ph.D. degree in computer science from the
[12] J. A. Al-Lawati and J. Tuomilehto, “Diabetes risk score in Oman: A tool University of Queensland, Brisbane, Australia.
to identify prevalent type 2 diabetes among Arabs of the middle east,” She is currently an Associate Professor in computer science with German
Diabetes Res. Clin. Pract., vol. 77, no. 3, pp. 438–444, 2007. University of Technology in Oman, Muscat, Oman. For more than ten years, she
[13] K. Morik, P. Brockhausen, and T. Joachims, “Combining statistical learn- has industry experience in the area of IT in a multinational environment. Her
ing with knowledge-based approach-A case study in intensive care moni- current research interests include machine learning and medical data mining.
toring,” in Proc. Eur. Conf. Mach. Learn., 1998, pp. 268–277.
[14] R. L. Ye and P. E. Johnson, “The impact of explanation facilities on user
acceptance of expert systems advise,” MIS Q., vol. 19, pp. 157–172, Jun.
1995.
[15] Z. Chen, J. Li, and L. Wei, “A multiple kernel support vector machine
scheme for feature selection and rule extraction from gene expression Andrew P. Bradley (SM’97) received the Ph.D. degree from the University of
data of cancer tissue,” Artif. Intell. Med., vol. 41, pp. 161–175, 2007. Queensland, Brisbane, Australia, in 1996.
[16] C. J. Wyatt and D. G. Altman, “Prognostic models: Clinically useful or Since 1996, he has been a Researcher in Australia, the United Kingdom,
quickly forgotten?” Brit. Med. J., vol. 311, pp. 1539–1541, 1995. and Canada. He is currently an Associate Professor in biomedical engineering
[17] N. Barakat, “Rule extraction from support vector machines: Medical di- with the University of Queensland. His research interests include biomedical
agnosis prediction and explanation,” Ph.D. thesis, School Inf. Technol. applications of pattern recognition in signal and image analysis.
Electr. Eng. (ITEE), Univ. Queensland, Brisbane, Australia, 2007. Dr. Bradley is a Chartered Electrical Engineer of the IET.
[18] N. Barakat and A. P. Bradley, “Rule extraction from support vector ma-
chines: A sequential covering approach,” IEEE Trans. Knowl. Data Eng.,
vol. 19, no. 6, pp. 729–741, Jun. 2007.
[19] N. Barakat and A. P. Bradley, “Rule extraction from support vector ma-
chines: Measuring the explanation capability using the area under the ROC Mohamed Nabil H. Barakat received the B.Sc. and M.Sc. degrees in internal
curve,” presented at the 18th Int. Conf. Pattern Recognit. (ICPR 2006), medicine from Alexandria University, Alexandria, Egypt and the M.Sc. degree
Hong Kong. in nephrology and metabolism from Cairo University, Cairo, Egypt.
[20] J. R. Quinlan, C4.5: Programs for Machine Learning. SanMateo, CA: He is a Consultant Physician and a Diabetologist with the Department
Morgan Kaufmann, 1993. of Noncommunicable Disease Surveillance and Control, Ministry of Health,
[21] M. Asfour, A. Lambourne, A. Soliman, S. Al-Behlani, D. Al-Asfoor, Muscat, Oman. He has a long clinical experience in treatment and management
and A. Bold, “High prevalence of diabetes mellitus and impaired glucose protocols of diabetes. He is involved in the national program of diabetes control
tolerance in the Sultanate of Oman: Results of the 1991 national survey,” with the Ministry of Health. His current research interests include diabetes and
Diabetic Med., vol. 12, pp. 1122–1125, 1995. cardiovascular diseases prevention and control techniques.