![](bg4.png)
Table 2: Comparison of Machine Learning Algorithms.
Algorithm Accuracy Precision Recall AUC
DT 73,77% 73,08% 67,86% 0,695
RF 77,05% 73,33% 78,57% 0,866
SVM 81,97% 90,48% 67,86% 0,903
NN 75.41% 72.41% 75,00% 0,864
NB 80,33% 75.00% 85,71% 0,886
Figure 3: Classification Algorithm’s Performance Parame-
ters.
Figure 3 shows that the Support Vector Machine
algorithm obtains the highest accuracy and highest
precision with an accuracy of 81.97% and precision of
90.48% then for the highest recall obtained by Na
¨
ıve
Bayes with a recall of 85.71% and the highest AUC
obtained by Support Vector Machine with a value of
0.903.
The cost parameter, commonly called C, works
as an SVM optimization to avoid misclassification in
each sample in the training dataset. The SVM algo-
rithm tries to reduce misclassification as much as pos-
sible when the value of C is too large. This will lead
to a loss of generalization properties of the classifier
(algorithm). Simply put, if C is too large, the decision
boundary becomes very small.
When the value of C is too small, misclassifica-
tion of data points will occur due to a wider decision
boundary. The wider decision boundary generalizes
well on both training and testing data but may clas-
sify some records incorrectly.
The C parameter in Support Vector Machine de-
termines the margin density between support vectors.
The greater the C value, the closer the margin density.
Testing uses the C parameter with values of 0.0, 1.0,
2.0, 3.0, 4.0, and 5.0 to determine the highest accu-
racy in the Support Vector Machine algorithm.
Based on table 3, the Support Vector Machine al-
gorithm with parameter C value of 3.0 produces the
best accuracy of 85.25%, a precision of 91.30%, a re-
call of 75.00%, and an AUC of 0.900, indicating that
the Support Vector Machine is an excellent classifica-
tion in predicting heart disease.
Table 3: Testing of Parameter C SVM.
C Accuracy Precision Recall AUC
0.0 81,97% 90,48% 67,86% 0,903
1.0 83,61% 90,91% 71,43% 0,908
2.0 81,97% 86,96% 71,43% 0,892
3.0 85,25% 91,30% 75,00% 0,900
4.0 81,97% 90,48% 67,86% 0,889
5.0 85,25% 91,30% 75,00% 0,896
4 CONCLUSIONS
This research uses a decision tree, random forest, sup-
port vector machine, neural network, and na
¨
ıve bayes,
and several machine learning algorithms were eval-
uated for their performance in heart disease classifi-
cation. The results indicated that the support vector
machine (SVM) had the highest accuracy and per-
formed the best compared to other algorithms. Then
testing uses the C parameter to determine the highest
accuracy in the Support Vector Machine algorithm. It
can be concluded that the higher the value of C, the
less likely the error in determining the solution. Con-
versely, the lower the value of C, the higher the pro-
portion of errors in determining the solution. Thus,
it is suggested to find the optimal C value. So the
Support Vector Machine algorithm with a parameter
C value of 3.0 produces the best accuracy. Therefore,
it can be concluded that SVM is the best algorithm
for heart disease classification among the ones tested
in this study.
REFERENCES
Ahmad, I. and Raja, N. (2019). Comparison of machine
learning algorithms for predicting cardiovascular dis-
eases. J Health Inform, 9:10–18,.
Alaa, A., Eltawil, A., and Aljarah, I. (2017). Heart disease
diagnosis using data mining techniques. J Med Syst,
41(1):1–9,.
Alabdulmohsin, A. and Alhazmi, M. (2021). Machine
learning approaches for cardiovascular disease predic-
tion: A comparative study. J Med Syst, 45(5):567,.
Aljarah, I., Al-Zoubi, O., Al-Lawati, A., and Al-Omari, M.
(2011). Heart disease prediction system using data
mining techniques. Int J Comput Appl, 18(1):39–44,.
Allah, E., El-Matary, D., Eid, E., and Dien, A. (2022). Per-
formance comparison of various machine learning ap-
proaches to identify the best one in predicting heart
disease. Journal of Computer and Communications,
10(02):1–18,.
Balakrishnan, S., Bhargavi, R., and Suganya, R. (2019).
Classification of heart disease using machine learning
techniques. Int J Eng Adv Technol, 8(2):663–667,.
ICAISD 2023 - International Conference on Advanced Information Scientific Development
8