ISSN(Online): 2320-9801

ISSN (Print): 2320-9798

International Journal of Innovative Research in Computer

and Communication Engineering
(An ISO 3297: 2007 Certified Organization)

Vol. 4, Issue 9, September 2016

A Survey on Identification of Diabetes Risk

Using Machine Learning Approaches
B.Senthil Kumar1, Sreejith.R2
Assistant Professor, Department of Computer Science, Sree Narayana Guru College, K.G.Chavadi, Coimbatore,
Tamil Nadu, India1
M.Phil Scholar, Department of Computer Science, Sree Narayana Guru College, K.G.Chavadi, Coimbatore,
Tamil Nadu, India2

ABSTRACT: Several health oriented studies used machine learning approaches for analysis, detection and prediction
of health risks from different attributes of patient health records. Diabetes is one of the common and wide spread health
issues in India. Diabetes mellitus type 2 or type2 diabetes is a long term metabolic disorder that is considered
by high insulin defiance, lack of insulin and high blood sugar levels. Several machine learning approaches such as
supervised learning, clustering and regression etc., have been proposed. This paper surveys different data mining
approaches used to handle health care information’s along with the result summary. This survey explores the popular
and effective machine learning techniques along with its pros and cons.

KEYWORDS: Diabetes, Machine Learning, data mining, classification and prediction.

I. INTRODUCTION

Data mining is a key role in the intelligent health domain [1]. There are several software’s and tools have been
used to diagnose and classifies health information’s based on the attributes. The huge size databases are included into
this process as input. This process resulted in data collection complication. The followings are the basic information’s
about the diabetes and its basic causes and symptoms.
Diabetes risk Prediction Model can support medical professionals and practitioners in predicting risk status
based on the clinical data records. In biomedical field data mining and its techniques plays an essential role for
prediction and analyzing different type of health issues. The healthcare industry gives huge amounts of healthcare data
and that need to be mined to ascertain hidden information for valuable decision selection. Determining hidden patterns
and relationships may often very tough and unreliable. The health record is classified and predicted if they have the
symptoms of Diabetes risk and using risk factors of disease [2]. It is indispensable to find the best fit algorithm that has
greater accuracy, speedy and memory utilization on prediction in the case of Diabetes.

A. DIABETES:
Diabetes is classified into three types:
 Type 1 Diabetes: It is a chronic condition in which the pancreas produces little or no insulin. This type of
Diabetes results from the pancreas's failure to produce enough insulin. This necessitates the individual to insert
insulin or carry an insulin pump. This form was previously referred to as "insulin-dependent diabetes mellitus"
(IDDM). The cause of type 1 diabetes is unknown [3].
 Type 2 Diabetes: begins with insulin resistance, a condition in which cells fail to respond to insulin properly.
As the disease progresses a lack of insulin may also develop. This form was previously referred to as "non
insulin-dependent diabetes mellitus" (NIDDM) or "adult-onset diabetes". The primary cause is excessive body
weight and not enough exercise.
 Gestational diabetes is the third main form and occurs when pregnant women without a previous history of
diabetes develop high blood-sugar levels.
as a consequence of the human bodies malfunction to generate insulin, and necessitates the individual to insert
insulin or carry an insulin pump. This category was previously indicated as “Insulin-Dependent Diabetes Mellitus”

Copyright to IJIRCCE DOI: 10.15680/IJIRCCE.2016. 0409135 16752

 ISSN(Online): 2320-9801
ISSN (Print): 2320-9798

International Journal of Innovative Research in Computer

and Communication Engineering
(An ISO 3297: 2007 Certified Organization)

Vol. 4, Issue 9, September 2016

(IDDM). The second category of DM is recognized as “Type II DM” as a consequence of insulin confrontation, a
situation in which cells are ineffective to exploit insulin appropriately, occasionally merged with an absolute insulin
insufficiency. This category also called as “Non Insulin Dependent Diabetes Mellitus” (NIDDM) or “adult-onset
diabetes”. At last, “gestational diabetes” takes place when conceived women without an earlier

There are several researches conducted epidemiological and public health studies regarding the associations
between anthropometric measurements and type2 diabetes. In general the diabetes detection, heart disease risk
detection are handled using effective data mining algorithms. The common steps to detect the risk of disease are
represented in fig 1.0.

Health
dataset Pre-processing Feature Feature Predicted
Features
extraction analysis grouping Result

.
Fig 1.0 steps involved in the health risk prediction process

The above fig 1.0 represents the basic process involved in risk prediction in health dataset. In order to predict the risk
with its features, this is important to find the pertinent and appropriate features from the health dataset. The above
process shows the feature selection and feature based risk calculation for diabetes and other health dataset.

II. DATA MINING IN HEALTHCARE MANAGEMENT

To aid healthcare organization and decision making, data mining applications can be developed to improved
classification and track risk of several diseases from patient health records, this types of design, analysis and decision
making processes reduces the many work in real time. Consider, in order to create enhanced diagnosis and treatment
procedures, this is quite tough when comparing the decision making from scientific literature. In this case, data mining
is used to avoid data analysis and clinical problems by managing effective healthcare datasets. Data mining can be used
to analyse vast amount of data and statistics to search for patterns [4]. This survey brings the tools and techniques used
in the health care management.
In this chapter, we provide a detailed study about the traditional data mining technique in health care
application, such as heart disease analysis, type classification, and diabetes and heart risk assessment.

Data mining techniques for Heart Disease:

Healthcare data clustering is the process of segmenting text Healthcare data’s into different groups based on
its similarity level. The clustering is the unsupervised learning process, where it won’t need any training samples for
the grouping process [5]. The followings are the popular clustering algorithms are used for Healthcare data
management.
There are n number of studies was carried out the heart disease identification and risk prediction using data
mining. From the statistical data, the risk factors are associated and from those associations, the risks are detected. The
factors such as patient’s age, gender, blood pressure, food habits, cholesterol, heredity and hypertension etc., the heart
disease details are collected and stored as huge training sample, from the set of data, the useful information’s are
extracted. In this paper [6] author detected useful patterns from the database and find the risk of heart disease. These
works are focused on classification process.

The classification algorithm such as Decision tree [7], naïve bayes[8], SVM [9], and neural network
algorithms[10] are included. And some bagging [] and semi supervised classifiers also used to detect heart disease risk.

Copyright to IJIRCCE DOI: 10.15680/IJIRCCE.2016. 0409135 16753

 ISSN(Online): 2320-9801
ISSN (Print): 2320-9798

International Journal of Innovative Research in Computer

and Communication Engineering
(An ISO 3297: 2007 Certified Organization)

Vol. 4, Issue 9, September 2016

Table 1.0 data mining algorithms are proposed for heart disease

Algorithm Description Papers used the algorithm Results

Decision Tree It’s a tree like graph model Tu, et al., The accuracy gained in this
used for classification. 2009 (J4.8 Decision Tree) paper is 78.9%.

Andreeva, P. 2006
Palaniappan, et al. 2007
SVM Support vector machine is a Kangwanariyakul, et 74.9%
fully supervised learning al. 2010 (linear support
process. It is a cost vector
effective and suitable for machine)
massive data
polynomial support 70.59%
vector machine

radial basis function 60.89%

kernel support vector
machine
Naïve Naive bayes is completely Srinivas, et al. 2010 84.14%
Bayes based on the training Naïve Bayes and
samples. This is a One Dependency
successful classifier when Augmented Naïve
the training data is huge. Bayes classifier

Neural Network Neural network is Kangwanariyakul, et 76.5%

dependent with neural al. 2010
schemas, and it divided
into linear, probabilistic,
radial and polynomial
based approaches.
Bagging Bagging is the process of Tu et al., 81.43%
iterative classification, 2009
where data’s are divided
and used different classifier

The table 1.0 shows the list of algorithms used to handle the heart disease. The table gives the basic
description, paper details along with its result. The results are given in accuracy. The accuracy has been collected from
the necessary paper result. The above table clearly shows the basic neural network and SVM can only give limited
percentage of accuracy than others. The highest accuracy is bagging method. So the studies in the heart disease and
diabetes can be extended from the above result. Several papers used different set of attributes, so the results may vary
according to the dataset.

Data mining techniques for Diabetes:

As like the heart disease, the diabetes and risk is classified and predicted by various data mining techniques in
the literature such as regression [11], decision tree[12], and Artificial neural networks (ANN)[13].

Algorithm Description Papers used the algorithm Results

ANN (Artificial neural This follows evolutionary Lee SM, Kang JO 2004 The accuracy gained in this
Network) process. paper is 73.52 %

Copyright to IJIRCCE DOI: 10.15680/IJIRCCE.2016. 0409135 16754

 ISSN(Online): 2320-9801
ISSN (Print): 2320-9798

International Journal of Innovative Research in Computer

and Communication Engineering
(An ISO 3297: 2007 Certified Organization)

Vol. 4, Issue 9, September 2016

Decision Tree (C5.0) It’s a most recent version Hu FB, Manson JE, 78.27%
of decision tree, which has
high accuracy and ability to
handle missing and null
values.
C5.0 follows the post-
pruning
approach, which removes
branches from a fully
grown tree.
Regression It’s a non linear regression Lai CL, Lai CL, Chien SW, 72.74%
method for predicting Fang K. 2007
diabetes risk. The main
advantage of using this is,
it supports categorical data.
Table 2.0 data mining algorithms are proposed for Diabetes disease

The table 2.0 shows the list of algorithms used to handle the diabetes. Because detection of diabetes is
necessary, it generates several diseases such as retinopathy, neurological disorders, eye related issues and stroke etc.,
The table gives the basic description, paper details along with its result. The results are given in accuracy. The accuracy
has been collected from the necessary paper result. The table 2.0 clearly shows the basic ANN and regression can only
give limited percentage of accuracy than Decision Tree. The highest accuracy is bagging method.

In paper [14], author compared three prediction models for diabetes using 12 different attributes. The authors
have taken C5.0 decision tree algorithm, ANN and regression algorithms. The results from the paper show that the
C5.0decision tree model performed best on classification accuracy. This paper concluded and suggests future assist
with optimal predictive models. This also proofs the result accuracy varies when the attributes are effectively utilized.

The paper [15] presents the knowledge about the diabetes from the web data such as like MEDLINE. The use
of text mining, the reviews and the web data’s are mined effectively. Form this, the research outlined the brief
knowledge about the diabetes.

SVM also used for diabetes disease detection, in paper [16] proposed supervised learning method, i.e SVM, to
improve the terms’ discriminating power for disease detection task. This paper utilizes vector space model for text
representation, this transforms the content of a Healthcare data into a hyper pane. In this study, the authors investigated
numerous unsupervised and supervised classification methods with SVM and ANN algorithms. Finally the paper shows
the supervised term weighting methods are good in performance.

VI. CONCLUSION

With the use of data mining approach, health data management grows tremendously. In the modern scenario,
diabetes is a common disease spreads to all, and several techniques were implemented with several common and few
uncommon features. With the numerous sizes in digital Healthcare data processes, the classification and prediction
based on the statistical data is very tough. In such environment effective factors should be identified and that should be
used for further disease analysis otherwise any type classification and prediction process is became ineffective.
this paper shows the advantages and disadvantages of several traditional classification algorithms based on
different techniques. However, the techniques almost concentrated on general classification process, where the
Healthcare data prediciton needs additional concentration and work to improve the following problem. The first
problem is discovering appropriate features with less effort and validation on discovered features are not yet studied.
And there is a need for a new system to handle the above problem in Healthcare data management disease risk
prediction.

Copyright to IJIRCCE DOI: 10.15680/IJIRCCE.2016. 0409135 16755

 ISSN(Online): 2320-9801
ISSN (Print): 2320-9798

International Journal of Innovative Research in Computer

and Communication Engineering
(An ISO 3297: 2007 Certified Organization)

Vol. 4, Issue 9, September 2016

REFERENCES

[1]. Han, j. and M. Kamber, Data Mining Concepts and Techniques. 2006: Morgan Kaufmann Publishers. 2. Lee, I.-N., S.-C. Liao, and M.
Embrechts, Data mining techniques applied to medical information. Med. inform, 2000.
[2]. Obenshain, M.K., Application of Data Mining Techniques to Healthcare Data. Infection Control and Hospital Epidemiology, 2004.
[3]. Sandhya, J., et al., Classification of Neurodegenerative Disorders Based on Major Risk Factors Employing Machine Learning Techniques.
International Journal of Engineering and Technology, 2010. Vol.2, No.4.
[4]. Thuraisingham, B., A Primer for Understanding and Applying Data Mining. IT Professional IEEE, 2000. 6. Ashby, D. and A. Smith, The Best
Medicine? Plus Magazine - Living Mathematics., 2005.
[5]. Liao, S.-C. and I.-N. Lee, Appropriate medical data categorization for data mining classification techniques. MED. INFORM., 2002. Vol. 27,
no. 1, 59–67, .
[6]. Ruben, D.C.J., Data Mining in Healthcare: Current Applications and Issues. 2009.
[7]. Porter, T. and B. Green, Identifying Diabetic Patients: A Data Mining Approach. Americas Conference on Information Systems, 2009.
[8]. Panzarasa, S., et al., Data mining techniques for analyzing stroke care processes. Proceedings of the 13th World Congress on Medical
Informatics, 2010.
[9]. Li L, T.H., Wu Z, Gong J, Gruidl M, Zou J, Tockman M, Clark RA, Data mining techniques for cancer detection using serum proteomic
profiling. Artificial Intelligence in Medicine, Elsevier, 2004.
[10]. Das, R., I. Turkoglu, and A. Sengur, Effective diagnosis of heart disease through neural networks ensembles. Expert Systems with
Applications, Elsevier, 2009. 36 (2009): p. 7675–7680.
[11]. World Health Organization. 2007 7-Febuary 2011]; Available from: http://www.who.int/mediacentre/factsheets/fs310.pdf.

Copyright to IJIRCCE DOI: 10.15680/IJIRCCE.2016. 0409135 16756