44
International Journal of Computer Applications (0975 888)
Volume 47 No.10, June 2012
The prediction of Heart disease, Blood Pressure and 1 age Age in years Continuous
Sugar with the aid of neural networks was proposed 2 sex Male or female 1 = male
by Niti Guru et al. [4]. The dataset contains records 0 = female
with 13 attributes in each record. The supervised 3 cp Chest pain type 1 = typical
networks i.e. Neural Network with back propagation type 1
algorithm is used for training and testing of data. 2 = typical
The problem of identifying constrained association type agina
rules for heart disease prediction was studied by 3 = non-agina
Carlos Ordonez [7]. The resultant dataset contains pain
records of patients having heart disease. Three 4=
constraints were introduced to decrease the number asymptomatic
of patterns [6]. They are as follows: 4 thestbps Resting blood Continuous
1. The attributes have to appear on only one side pressure value in mm
of the rule. hg
2. Separate the attributes into groups. 5 chol Serum Continuous
i.e. uninteresting groups. cholesterol value in
3. In a rule, there should be limited number of mm/dl
attributes. 6 Restecg Resting 0 = normal
The result of this is two groups of rules, the electrographic 1=
presence or absence of heart disease. results having_ST_T
wave
Franck Le Duff et al. [9] builds a decision tree with abnormal
database of patient for a medical problem. 2 = left
Latha Parthiban et al. [10] projected an approach on ventricular
basis of coactive neuro-fuzzy inference system hypertrophy
(CANFIS) for prediction of heart disease. The 7 fbs Fasting blood 1 120 mg/dl
CANFIS model uses neural network capabilities sugar 0 120 mg/dl
with the fuzzy logic and genetic algorithm. 8 thalach Maximum heart Continuous
Kiyong Noh et al. [8] uses a classification method rate achieved value
for the extraction of multiparametric features by 9 exang Exercise 0= no
assessing HRV (Heart Rate Variability) from induced agina 1 = yes
ECG, data pre-processing and heart disease pattern. 10 oldpeak ST depression Continuous
The dataset consisting of 670 peoples, distributed induced by value
into two groups, namely normal people and patients exercise relative
with heart disease, were employed to carry out the to rest
experiment for the associative classifier. 11 solpe Slope of the 1 = unsloping
peak exercise 2 = flat
ST segment 3=
4. PROPOSED PREDICTION SYSTEM downsloping
Today, many hospitals manage healthcare data using 12 ca Number of 0-3 value
healthcare information system; as the system contains huge major vessels
amount of data, used to extract hidden information for making colored by
intelligent medical diagnosis. The main objective of this floursopy
research is to build Intelligent Heart Disease Prediction 13 thal Defect type 3 = normal
System that gives diagnosis of heart disease using historical 6 = fixed
heart database. To develop this system, medical terms such as 7 = reversible
sex, blood pressure, and cholesterol like 13 input attributes are defect
used. To get more appropriate results, two more attributes i.e.
obesity and smoking are used, as these attributes are
considered as important attributes for heart disease. The data All the research papers referred above have used 13 input
mining classification techniques viz. Neural Networks, attributes for prediction of Heart disease. To get more
Decision Trees, and Naive Bayes are used. appropriate results two important attributes i.e. obesity and
smoking are added to input attributes.
5. DATA SOURCE Table 2. Description of newly added attributes
The publicly available heart disease database is used. The
Cleveland Heart Disease database [11] consists of 303 records
& Statlog Heart Disease database consists of 270 records [12]. Sr. Attribute Description Values
The data set consists of 3 types of attributes: Input, Key & no
Predictable attribute which are listed below. 14 obes obesity 1 = yes
0 = no
5.1. Input attributes 15 smoke smoking 1= past
2 = current
Table 1. Description of 13 input attributes 3 = never
45
International Journal of Computer Applications (0975 888)
Volume 47 No.10, June 2012
It maps a set of input data onto a set of appropriate output The Bayes theorem is as follows:
data.It consists of 3 layers input layer, hidden layer & output
layer. There is connection between each layer & weights are Let X={x1, x2, ....., xn} be a set of n attributes. In Bayesian, X
assigned to each connection. The primary function of neurons is considered as evidence and H be some hypothesis means,
of input layer is to divide input xi into neurons in hidden layer. the data of X belongs to specific class C. We have to
Neuron of hidden layer adds input signal xi with weights wji of determine P (H|X), the probability that the hypothesis H holds
respective connections from input layer. The output Yj is given evidence i.e. data sample X. According to Bayes
function of theorem the P (H|X) is expressed as
46
International Journal of Computer Applications (0975 888)
Volume 47 No.10, June 2012
Table 3. A confusion matrix Table 3 shows accuracy for different classification methods
with 13 input attributes & 15 input attributes values.
a (has heart b (no heart
disease) disease) Table 6. Comparison of data mining techniques
47
International Journal of Computer Applications (0975 888)
Volume 47 No.10, June 2012
[3] Sellappan Palaniappan, Rafiah Awang, "Intelligent [8] Kiyong Noh, Heon Gyu Lee, Ho-Sun Shon, Bum Ju
Heart Disease Prediction System Using Data Lee, and Keun Ho Ryu, "Associative Classification
Mining Techniques", IJCSNS International Journal Approach for Diagnosing Cardiovascular Disease",
of Computer Science and Network Security, Vol.8 Springer, Vol:345, pp: 721- 727, 2006.
No.8, August 2008
[9] Franck Le Duff, Cristian Munteanb, Marc Cuggiaa,
[4] Niti Guru, Anil Dahiya, Navin Rajpal, "Decision Philippe Mabob, "Predicting Survival Causes After
Support System for Heart Disease Diagnosis Using Out of Hospital Cardiac Arrest using Data Mining
Neural Network", Delhi Business Review, Vol. 8, Method", Studies in health technology and
No. 1 (January - June 2007). informatics, Vol. 107, No. Pt 2, pp. 1256-9, 2004.
[5] Heon Gyu Lee, Ki Yong Noh, Keun Ho Ryu, [10] Latha Parthiban and R.Subramanian, "Intelligent
Mining Biosignal Data: Coronary Artery Disease Heart Disease Prediction System using CANFIS and
Diagnosis using Linear and Nonlinear Features of Genetic Algorithm", International Journal of
HRV, LNAI 4819: Emerging Technologies in Biological, Biomedical and Medical Sciences, Vol.
Knowledge Discovery and Data Mining, pp. 56-66, 3, No. 3, 2008.
May 2007.
[11] Cleveland database:
[6] Shantakumar B.Patil, Y.S.Kumaraswamy http://archive.ics.uci.edu/ml/datasets/Heart+Disease
Intelligent and Effective Heart Attack Prediction
System Using Data Mining and Artificial Neural [12] Statlog database:
Network. ISSN 1450-216X Vol.31 No.4 (2009), http://archive.ics.uci.edu/ml/machine-learning-
pp.642-656. databases/statlog/heart/
[7] Carlos Ordonez, "Improving Heart Disease [13] Dr. Yashpal Singh, Alok Singh chauhan Neural
Prediction Using Constrained Association Rules," Networks in data mining Journal of Theoretical
Seminar Presentation at University of Tokyo, 2004. and Applied Information Technology , 2005 - 2009
JATIT.
48