Page 56 - AIH-1-1

Artificial Intelligence in Health AI model for cardiovascular disease prediction

and precision were observed to increase as the number accuracy and sensitivity decreased but increased when 8
of attributes reduced from 14 to 10 and 8. However, the was used, whereas the MSE and specificity increased and
accuracy and sensitivity decreased as the number of slightly decreased when the number of attributes was
attributes reduced from 14 to 10 and then increased as the reduced to 8. The details of the selected attributes for 14,
number of attributes was further reduced to 8. The results 10, and 8 are presented in Tables 4-6.
demonstrated that the system performed better with 14 4.2. Performance evaluation analysis of
attributes, compared to other selected attributes such as 10
and 8, as illustrated in Figure 9. classification algorithm on the varied number of
datasets
The performance analysis of using DT algorithm on The number of the dataset was varied from 2000, 1500,
the selected attributes of 14, 10, and 8 was conducted. 1000, and 500. The corresponding performance on
The results obtained for the accuracy, MSE, sensitivity, different classifiers was observed. The performance analysis
specificity, and precision were 77.80%, 0.222, 91.67%, result of various models using 2000 sets of data with ANN
57.6%, and 75.9%, respectively. The results presented is presented in Figure 11. The results obtained for the
in Figure 10 showed that a better result was achieved accuracy, MSE, sensitivity, specificity, and precision were
when 14 attributes were selected. But when the selected 73.30%, 0.2667, 77.80%, 66.70%, and 77.80%, respectively.
attributes were reduced to 10, the classification accuracy, The ANN classification showed better performance with a
MSE, sensitivity, specificity, and precision obtained were high range of datasets; hence, 2000 datasets were adopted.
70.4%, 0.296, 77.10%, 60.60%, and 74.0%, respectively. When the cardiac disease dataset was reduced to 1500,
For the selected 8 attributes, the obtained accuracy, MSE, the accuracy, MSE, sensitivity, specificity, and precision
sensitivity, specificity, and precision were 77.80%, 0.222, obtained were 71.10%, 0.2889, 82.60%, 59.10%, and
97.9%, 50.00%, and 73.0%, respectively. However, when 67.90%, respectively. Considering 1000 CVD datasets, the
the number of attributes was reduced from 14 to 10, the results obtained for accuracy, MSE, sensitivity, specificity,
and precision were 83.30%, 0.1667, 90.00%, 70.00%, and
85.70%, respectively. Finally, for the 500 CVD datasets,
the accuracy, MSE, sensitivity, specificity, and precision
results obtained were 66.70%, 0.3330, 83.30%, 55.60%,
and 66.70%, respectively. However, when the selected
CVD dataset number was reduced from 2000 to 1500, the
accuracy, specificity, and precision decreased and increased
sharply at 1000 datasets before decreasing again when
the number of cardiac dataset was 500. The sensitivity
increased when the dataset was reduced from 2000 to 1000
and decreased only slightly when the dataset number was
500. Meanwhile, the MSE increased at 1500 CVD datasets
using ANN, decreased when the dataset used was 1000,
Figure 8. Performance analysis of the K-nearest neighbor algorithm on and increased slightly when the number of cardiac dataset
the selected attributes. dropped to 500.

Figure 9. Performance analysis of support vector machine algorithm on Figure 10. Performance analysis of using decision tree algorithm on the
the selected attributes. selected different attributes.

Volume 1 Issue 1 (2024) 50 https://doi.org/10.36922/aih.1746