International Journal of Innovative Technology and Exploring Engineering (IJITEE)

ISSN: 2278-3075 (Online), Volume-9 Issue-3, January 2020

Heart Diseases Prediction using Deep Learning

Neural Network Model
Sumit Sharma, Mahesh Parmar
 Keras is a deep learning neural network library. Keras
Abstract:- Deep learning plays an important role in the field creates a high-level neural networks model. It developed for
of medical science in solving health issues and diagnosing various easy and fast experimentation. Keras supports both
diseases. So in this paper, we will discuss heart disease. convolutional neural networks (CNN) and recurrent neural
We proposed a model for heart disease prediction. Heart
networks (RNN), as well as combinations of the two. It runs
Disease is on of key area where Deep Neural Network can be used
so we can improve the overall quality of the classification of heart perfectly on CPU and GPU. Talos is a fully automated POD
disease. The classification can be performed on the various ways (Prepare, Optimize, Deploy) pipeline that stability yields state
like KNN, SVM, Naïve Bayes, Random Forest. Heart Disease UCI by step prediction results in a wide range of prediction related
dataset will be used to demonstrate Talos Hyper-parameter problems. In this field, large amount of researches has been
optimization is more efficient than others. done by using various algorithms and techniques.
This paper aims to achieve better accuracy and to make
Keyword:- Deep Learning, Neural Network (NN), CNN, the system more efficient so that it can predict the chances of
RNN, KNN, SVM, Heart disease dataset.
a heart attack.
I. INTRODUCTION
II. RELATED WORK
In our day to day life,Today generation is very busy in our
Nowadays many researchers used machine learning,
daily routine schedule which feels to nervousness,
deep learning and data mining in healthcare for the predict to
restlessness and stress.Each individual have distinctive pulse
the disease but each research gives your opinion and your
rate and blood pressure which ranges from 60 to 100 BPM for
prediction accuracy according to your research.
pulse rate and 120/80 to 140/90 for blood pressure.
Tan et al. [1] suggested a hybrid method using two
In world wide, heart disease is the major issue in human
machine learning algorithms, one is SVM (Support Vector
life. Heart means “Cardio”. Heart disease category is called
Machine) and another is GA (Genetic Algorithm), both are
Cardiologist disease.
effectively combined with in this approach. Data mining tools
like LIBSVM and WEKA are used for this analysis here we
The different types of heart disease are following :-
collected 5 different dataset from the IUC repository. When
 Congenital heart disease.
we applied the hybrid model it reaches to an accuracy of
 Arrhythmia. 84.07% for heart disease, 78.26 for diabties and 76.20% for
 Coronary artery disease. breast cancer and 86.12% for Hepatitis
 Dilated cardiomyopathy. Chaurasia et al.[2] recommended data mining
 Myocardial infarction. perspectives to detected heart diseases in human body. In this
 Heart failure. data mining approaches, WEKA tool is used for machine
 Hypertrophic cardiomyopathy. learning purposes which is used multiple algorithms for data
 Mitral regurgitation mining like - J48, Naïve bayes and bagging. The UCI
laboratory is a part of machine learning. In the heart disease
Deep learning is a more popular machine learning method. dataset 313 attributes and 13 attributes for prediction. Naïve
It is not only when applying it in image classification tasks but Bayes gives 82.31% accuracy, J48 offers 84.35% and
also uses normal tabular data. In this model, we create a deep Bagging offers 85.35% accuracy in classification.
learning neural network model using Talos. As suggested by Vembandasamy et al. [3] using the
Talos is a hyperparameter optimization techniques. In this Naive Bayes algorithm which uses Bayes approach. Naïve
model, we can use the Talos optimizer with Keras library. Bayes algorithm has a robust principle of independence
dataset. Here we use one of the leading diabetes research
dataset which consist of 500 patients. WEKA tools is data
Revised Manuscript Received on January 30, 2020.
* Correspondence Author
mining tools and performs classification using 70% for
Sumit Sharma*, M. Tech. Scholar,Dept. of CSE & IT, MITS Gwalior, training and 30% for testing dataset. Accuracy of Naive
Gwalior, Madhya Pradesh, India. Email: s.s.sumit1996@gmail.com Bayes 86.419%.
Mahesh Parmar, Assistant Professor MITS,Dept. of CSE & IT, MITS SahayaArthyet.al[4]analyses the existing works on heart
Gwalior, Gwalior, Madhya Pradesh, India. Email:
disease prediction which uses data mining. The data mining
maheshparmarcse@mitsgwalior.in
techniques are commonly used in heart disease prediction.
© The Authors. Published by Blue Eyes Intelligence Engineering and They also discuss the databases used such as the heart disease
Sciences Publication (BEIESP). This is an open access article under the data set from UCI repository, tools used such as Weka, Rapid
CC-BY-NC-ND license http://creativecommons.org/licenses/by-nc-nd/4.0/ Miner, Data melt,

Published By:
Retrieval Number: C9009019320/2020©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijitee.C9009.019320 2244 & Sciences Publication
Journal Website: www.ijitee.org
 Heart Diseases Prediction using Deep Learning Neural Network Model

Apache Mahout, Rattle, KEEL, R data mining and soon. Type 2: non-anginal pain
They conclude that the use of single algorithm results in better Type 3: asymptomatic
4. Trestbps Resting Blood sugar (in mm Hg on
accuracy in prediction. But the use of hybridization of two or admission to the hospital).
more algorithms can enhance and improve the heart disease 5. Chol serum cholesterols in mg/dl
prediction with good accuracy. 6. Fbs Fasting blood suger > 120 mg/dl.
Maratea et al, [5] evaluated a sequential feature selection (1= true; 0= false)
approach using a neuro-fuzzy classifier. Its obtaining the 7. Restecg Resting ECG result
8. Thalach Maximum heart rates Achieved.
Cleveland set accuracy of 88.2%. in this paper, training data 9. Exang Exercise induced angina.
is using 50% of dataset and testing data is using 50% of the (1 = yes; 0 = no)
dataset. 10. Oldpeak ST depression induced by
Lakshmi et. al. [6] maintaining 10 methods, Its using exercise relative to rest.
heart disease UCI repository in the 4 sets, and found that a 11. Slope Slope or peak exercise ST Segment.
Value 1:upsloping
Partial Least Square Discriminant Analysis (PLS-DA) Value 2: flat
method accuracy of 86.13%. Value 3: downsloping
R. Sharmila et al, [7] as suggested to enhanced the 12. Ca number of major vessels (0-3)
prediction of heart diseases dataset using the data techniques. colored by flourosop
13. Thal 3 = normal;
SVM provided the better and efficient accuracy 85%. In 6 = fixed defect;
SVM, parallel fashion gives better accuracy than sequential 7 = reversable defect
SVM. 14. Num The predicted attribute.
Ashwini Shetty A et al, [9] Different Data Mining 0 : Yes;
Approaches for Predicting Heart Disease. WEKA tool, 1 : No.
MATLAB. Accuracy of Neural Network 84% and Accuracy
of Hybrid Systems 89%. B. Data cleaning: - Data cleaning is the first and necessary
Chala Bayen et al, [10] as suggested to Prediction and step for any project processes and data models. It means
Analysis for Heart Diseases using data mining techniques. filtering and modifying your data such that it is easier to
J48, Naïve Bayes (NB), Support Vector Machine (SVM). It explore and understand.
gives the better result which helps to improved the quality of
services and reduce the cost to individuals.

III. PROPOSED METHODOLOGY

In this paper, We deploy a model “Optimized DNN using
Talos” and compare the method to others it is more efficient
to others. This model provided a high accuracy compared to
others. In this model, we are following some steps.

Fig.2- Heat-map
C. Feature engineering:- Features engineering is the most
popular part of the deep learning. Features engineering is
used for extracting some features data set. Features
engineering is the process of transferring the raw data to
features data and this features data improve the quality of the
model and provide better accuracy.
Fig.1- Flow diagram D. Learning algorithms:- We are using the many types of
learning algorithms –
A. Dataset:- In this paper, we are using the Heart Diseases a) Logistic regression :- Logistic regression is a
datasets. In Heart Diseases, 14 attributes are given shown in classification algorithm for categorical variables. Sigmoid
fig.2 and 303 columns who represents the patient's data. In function is the most part of logistic regression.
this database, we apply the DNN using Talos and predict the Algorithms :- Logistic regression algo. following some
diseases. steps.
1. Initialize ϴ.
Table 1 – Dataset description
No. Attributes Descriptions 2. Calculate Y = σ(ϴTX) for a customer.
1. Age Patients Age (in Year) 3. Compare the output of Y with actual output of
2. Sex 0 : female and 1 : male customer y, record it as error.
3. Cp Type of Chest pain 4. Calculate the cost of all
Type 0: typical Angina customers.
Type 1: atypical angina

Published By:
Retrieval Number: C9009019320/2020©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijitee.C9009.019320 2245 & Sciences Publication
Journal Website: www.ijitee.org
 International Journal of Innovative Technology and Exploring Engineering (IJITEE)
ISSN: 2278-3075 (Online), Volume-9 Issue-3, January 2020

5. Change the ϴ to reduce the cost.

6. Go To Step 2.
7. Stop.

Fig.5- Support vector

The SVM algorithm is used to predict this disease by plotting

the training dataset where a hyperplane classifies in two –
presence and absence of heart disease.

Fig.3- Logistic regression

b) K-NN:- KNN is a non- parametric machine learning

algorithm. It is a supervised learning algorithm. It means
to predict the output from the input data.

Fig.6- SVM

SVM is used to handle class imbalance. Class imbalance

is a problem in machine learning when the total number of
positive and negative is not the same and the classifier will not
perform well
d) Naïve Bayes:- Naive Bayes is a probabilistic machine
learning classification algorithm based on the Bayes
Theorem. It is used in a wide variety of classification tasks.
Bayes Rule is a way of going from P(X|Y), known from
the training dataset, to find P(Y|X), known from the test data.

Fig.4- KNN

K-NN Algorithms:- K-NN algo following some steps.

1. Pick a value for K. P(X|Y) is known data means P(Evidence|Outcome) for testing
2. Calculate the distance of unknown case from all cases. and P(Y|X) is unknown data means P(Outcome|Evidence) for
3. Select the K- observations in the training data that nearest testing.
to the unknown data point.
4. Predict the response of unknown data point using the
most popular response value from the KNN
5. Stop.
e) Hyper-parameter optimization(Talos) :- Talos follows
In this algorithms data is divided into training and test data POD (Prepare, optimize, Deploy), process workflow and
sets. The training dataset is used for model building and additional functionality for evaluation, reporting, including
training. K- value is decided which is often the square root of plots for visual analysis.
the number of observations. Now the test data is predicated on Prepare (P): - this is the first process for preparation of
the model built. defining the hyperparameter space for the experiments and it
is the setting oof experiment options such as choosing for the
c) Support Vector Machine(SVM):- SVM is supervised deep optimization strategy.
learning algorithm which can be used both classification Optimize (O): - This is the second process for optimization,.
and regression. Its automated process of finding an optimal hyperparameter
combination for a well generalizing model for a given
prediction task.
Deploy (D): - This is the third process for deployment.

Published By:
Retrieval Number: C9009019320/2020©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijitee.C9009.019320 2246 & Sciences Publication
Journal Website: www.ijitee.org
 Heart Diseases Prediction using Deep Learning Neural Network Model

Its automated the process of sorting locally the required assets F. Evaluate model:-We deploy the neural networks model. In
for local and remote deployment of a model for production this model, we are using the input layer, output layer,
purpose. hidden layer, and activation function. In this model, we
Reporting and evaluation is the last process after all 3 POD deploy the deep neural network model using Talos
workflow. Its provides several facility for analysis and optimization. The purpose of Talos optimization is
evaluation of experiments, including the all plots for allowing to use to continue working with Keras models.
epoch-by-epoch visual analysis for experimental progress.

f) Random forest classifier :- Random forest classifier makes

a set of decision trees from randomly selected subset of
training dataset. It aggregates the votes from different
decision trees to decide the final class of the test object.
Random Forest Algorithms :- This algo follows the some
steps.
1. Choose an attributes from your dataset.
2. Calculate the significance of attribute in splitting of data.
3. Split data based on the value of the best attribute.
4. Go to Step 1.

E. Train & Test model:-Data is split into test and train dataset
where The train set contain a output field on which the
model learns.

Fig.7- Data splitting

Fig.10- Hyper-parameter optimization technique using
Here we applied cross-validation so our data remains Talos
randomly across the train set.
IV. RESULTS
In this paper we applied some classification algorithms
(like – K-NN, SVM, Hyper-parameter optimization ) on
Heart diseases data set and measure the all classification
accuracy is available on below mention table.

Table 2.- Algorithms accuracy comparison

S.No. Classification Algorithms Accuracy
1. Logistic Regression 85.25%
2. K-NN 90.16%
3. SVM 81.97%
4. Naïve Bayes 85.25%
5. Hyper-parameter optimization 90.78%
(Talos)
6. Random forest 85.15%
Fig.8 - Training to testing process
In this paper, we finally found out the best classification
In the training/testing model, two words are very popular first algorithms is Hyper-parameter optimization using Talos for
is Under-fitting and second is Over-fitting. Heart diseases dataset.
If the model is under-fitting then our model doesnot fill well
and perfor poorly
And if it Over-fits then it performs poorly on test set.

Fig.11- Algorithms Accuracy comparison

Fig.9 –Under-fitting and over-fitting

Published By:
Retrieval Number: C9009019320/2020©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijitee.C9009.019320 2247 & Sciences Publication
Journal Website: www.ijitee.org
 International Journal of Innovative Technology and Exploring Engineering (IJITEE)
ISSN: 2278-3075 (Online), Volume-9 Issue-3, January 2020

V. CONCLUSIONS
A summary of this paper arranged in alogical sequence
that generally follows your methodology section. Compare to
other algorithms and optimization, it is proved good results
for prediction. In this paper, We deploy a deep learning
neural networks (DNN) using Talos optimization. Talos
optimization is newly optimization techniques in DNN. Talos
provide better accuracy (90.76%) to other optimizations. It is
applied on the Heart disease datasets and find out the good
prediction. Using the Talos optimization we create a Keras
model and deploy it.

REFERENCES
1. K. C. Tan, E. J. Teoh, Q. Yu, and K. C. Goh, "A hybrid evolutionary
algorithm for attribute selection in data mining", Expert Systems with
Applications, Vol.36, No.4, pp.8616-8630, 2009.
2. V. Chaurasia and S. Pal, "Data mining approach to detect heart
diseases", International Journal of Advanced Computer Science and
Information Technology, Vol.2, No.4, pp.56-66, 2014.
3. K. Vembandasamy, R. Sasipriya, and E. Deepa, "Heart Diseases
Detection Using Naive Bayes Algorithm", IJISET-International
Journal of Innovative Science, Engineering & Technology, Vol.2,
pp.441-444, 2015.
4. A.Sahaya Arthy, G. Murugeshwari “A survey on heart disease
prediction using data mining techniques” (April 2018).
5. Hamid Reza Marateb and Sobhan Goudarzi, “A noninvasive method
for coronary artery diseases diagnosis using a clinically interpretable
fuzzy rule-based system,” Journal of Research in Medical Sciences,
Vol. 20, Issue 3, pp.214-223, March 2015.
6. K.R. Lakshmi, M.Veera Krishna, and S.Prem Kumar, “Performance
Comparison of Data Mining Techniques for Predicting of Heart
Disease Survivability,” International Journal of Scientific and
Research Publications, Vol.3, Issue 6, pp.1-10, June 2013.
7. R. Sharmila, S. Chellammal, “A conceptual method to enhance the
prediction of heart diseases using the data techniques”, International
Journal of Computer Science and Engineering, May 2018.
8. Purushottam, Prof. (Dr.) Kanak Saxena, Richa Sharma, “Efficient
Heart Disease Prediction System”, 2016, pp.962-969.
9. Ashwini Shetty A, Chandra Naik, “Different Data Mining
Approaches for Predicting Heart Disease”, International Journal of
Innovative in Science Engineering and Technology, Vol.5, May 2016,
pp.277-281.
10. Mr. Chala Beyene, Prof. Pooja Kamat, “Survey on Prediction and
Analysis the Occurrence of Heart Disease Using Data Mining
Techniques”, International Journal of Pure and Applied Mathematics,
2018.

AUTHORS PROFILE

SUMIT SHARMA, (M.tech.Scholar), Dept. of CSE &

IT, MITS Gwalior, Madhya Pradesh, India. Area of
Interest: Data Mining, Image Processing, Machine
Learning and Deep Learning. I am an student of CSE&IT
Department in MITS Gwalior. I have received B.E.
degree in information technology from MITS Gwalior. I
am three times Gate qualified 2017, 2018 and 2019.

Mr. Mahesh Parmar, as an Assistant Professor in

CSE&IT Department in MITS Gwalior and having 10
years of Academic and Professional experience. He
received M.E. degree in Computer Engineering from
SGSITS Indore. He has guided several students at
Master and Under Graduate level. His areas of current
research include Data mining and Image Processing. He
has published more than 30 research papers in the
journals and conferences of international repute. He has also published 02
book chapters. He is having the memberships of various Academic/
Scientific societies including IETE, CSI, and IET etc.

Published By:
Retrieval Number: C9009019320/2020©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijitee.C9009.019320 2248 & Sciences Publication
Journal Website: www.ijitee.org