978-1-5090-4815-1/17/$31.00 ©2017 IEEE
Evaluating Ensemble Prediction of
Coronary Heart Disease using Receiver
Operating Characteristics
Tahira Mahboob
1
, Rida Irfan
2
, Bazelah Ghaffar
3
Department of Software Engineering
Fatima Jinnah Women University, The Mall, Rawalpindi, Pakistan
tahira.mahboob@yahoo.com
Abstract- Heart diseases may perhaps consequence
in debility, severe disorder, and meager quality of
lifespan. Furthermore, it could also be lethal. Hence
inferring heart disease has turn into foremost distress
currently. This paper centers on various machine
learning practices which assist ascertaining and
perceiving innumerable heart diseases. Multifarious
machine learning approaches conversed here are
Hidden Markov Models, Support Vector Machine,
Feature Selection, Computational intelligent
classifier, prediction system, data mining techniques
and genetic algorithm. Scrutinizing each approach
thoroughly allowed us to select most apposite one.
This ultimately permits us to propose an Ensemble
Model exploiting pertinent machine learning
procedures which perfectly categorizes diverse heart
diseases. The evaluation of the proposed technique
has been conducted using state of the art technology.
The proposed technique has an accuracy of 94.21%,
a ROC (Receiver Operating Characteristics) of
0.981, RMSE (Root Mean Square Error) of .2568,
Precision of 0.953; showing significant
improvement when compared to the performance of
K-Nearest Neighbor, Artificial Neural Networks
and Support Vector Machines algorithms.
Analysis/Evaluation of the implemented algorithms
and the proposed Ensemble Model has been done
expending the Receiver Operator Characteristics.
Keywords- ANN (Artificial Neural Networks),
Ensemble Model KNN (K-Nearest Neighbor), SVM
(Support Vector Machines), ROC.
I. INTRODUCTION
In developed and under developing countries,
prominent origin of death is heart disease. Person’s
health is significantly influenced by the heart disease
suffered. Cardiovascular disease (CVD) is endured
by 80 000 000 inhabitant, alone in united states.
Each day approximately 2400 Americans die
because of this disease. One very common form of
CVD is Anomalous heart rhythm termed as cardiac
arrhythmia. The correct functionality of the heart is
significantly influenced by Cardiac Autonomic
Neuropathy (CAN). Deposits of fatty acids in
coronary artery may constrict it down and result in
coronary heart disease, which grounds for an
occurrence of 1.2 million heart attacks each year.
Providing eminent services is the major concern
faced by the health care administrations currently.
For instance, it requires early diagnosis of heart
disease efficiently and effectively. Hence in order to
accomplish this task we are executing various heart
disease prediction mechanisms followed by
proposing Ensemble models. As the irregular
heartbeats are easily perceived by
electrocardiogram, therefore ECG seems to be quiet
helpful for physicians particularly for the bulky
volumes of statistics. This Research paper is
systemized in following manner. Section-I is the
introduction. Section-II summarizes all the research
papers reviewed. Section-III converses the
implementation of Ensemble Model along with data
set and result analysis, followed by concluding the
research paper in fourth section.
II. LITERATURE REVIEW
Expending Artificial Neural Network for prediction
of heart disease is major focus of Wijaya et. al[1].
Moreover, Support Vector is also being considered
for the prediction process. Predicting heart
syndrome is possible within a year by overviewing
irregular heart rate. Utilizing various tools such as
smart mirror, smart chair, smart mouse and smart
phone, data regarding individuals is collected in a
server. This is how fatality rate along with number
of patients suffering from heart disease decrements
significantly. However, in Year 2011 observed
accuracy of ANN was 80.06% while SVM observed
accuracy was 84.12%. Chen et. al. in 2011[2]
present Diagnosis of heart syndrome depends upon
the medical data. Heart syndrome prediction system
is developed that can help the medical professionals
in predicting the heart syndrome by analyzing the
medical data of patients. The system takes thirteen
medical attributes as input. Then system uses the
ANN technique for categorizing the heart syndrome
on the basis of these medical attributes. Moreover,
ROC is displayed that depicts the performance of
110