Page 58 - AIH-1-1

Artificial Intelligence in Health AI model for cardiovascular disease prediction

Figure 14. Performance analysis of K-nearest neighbor algorithms on
Figure 12. Performance analysis of K-means algorithms on datasets of datasets of different sizes.
different sizes.
and precision were 54.00%, 0.4600, 100%, 100%, and 50%,
respectively. The results showed that KNN has excellent
sensitivity and specificity of 1000% even at a reduced
number of datasets. Accuracy and precision decreased
gradually as the number of datasets reduced from 2000
to 500. On the other hand, MSE increased steadily with a
reduced number of datasets.
The performance results of the model using the SVM
classification technique on 2000, 1500, 1000, and 500
datasets are presented in Figure 15. The classification
accuracy, MSE, sensitivity, specificity, and precision were
78.30%, 0.2170, 91.70%, 58.30%, and 96.10%, respectively,
for 2000 datasets. According to the results, the system
Figure 13. Performance analysis of artificial neural network-genetic performed better at 2000 datasets, hence adopted for the
algorithms on datasets of different sizes
system. When the dataset number was reduced to 1500, the
reduced from 2000 to 500. MSE increased with a reduced accuracy, MSE, sensitivity, specificity, and precision were
number of datasets. Specificity increased as the number of 71.10%, 0.2890, 84.00%, 55.00%, and 70.00%, respectively.
datasets was reduced from 2000 to 1500, decreased at 1000 For 1000 datasets, the accuracy, MSE, sensitivity, specificity,
CVD datasets, but increased sharply when the number and precision were 73.30%, 0.2670, 100%, 75.00%, and
of datasets was reduced to 500. Meanwhile, the precision 85.70%, respectively, and for the 500 datasets, the same
decreased as the number of datasets was reduced from set of performance parameters was measured at 58.30%,
2000 to 1500, increased at 1000 datasets, and decreased as 0.4170, 66.70%, 50.00%, and 57.10%, respectively. The
the number of datasets was reduced to 500. accuracy, sensitivity, specificity, and precision decreased
as the number of datasets reduced from 2000 to 1500,
The performance of the KNN classification technique then increased as the number of datasets was 1000 before
using 2000, 1500, 1000, and 500 datasets is presented in decreasing at 500 datasets. While the MSE increased when
Figure 14. The classification accuracy, MSE, sensitivity, datasets reduced to 1500, it decreased at 1000 datasets
specificity, and precision were 71.70%, 0.2838, 100%, before increasing to its peak at 500 datasets.
70.00%, and 67.90%, respectively, for 2000 datasets. Based
on the results, the system performed better at 2000 datasets The performance results of the system using the DT
when compared with 1500, 1000, and 500 datasets, and algorithm for 2000, 1500, 1000, and 500 datasets are
hence, 2000 datasets were adopted for the system. When the presented in Figure 16. For 2000 datasets, the classification
dataset number was reduced to 1500, the accuracy, MSE, accuracy, MSE, sensitivity, specificity, and precision were
sensitivity, specificity, and precision were 55.60%, 0.4440, 75.00%, 0.2500, 88.90%, 54.20%, and 74.40%, respectively.
100%, 100%, and 55.60%, respectively. For 1000 datasets, It was observed that the system performed better at 2000
the accuracy, MSE, sensitivity, specificity, and precision datasets; therefore, it was adopted for the system. Based on
were 56.70%, 0.4330, 100%, 100%, and 56.70%, respectively. the results, when the datasets were reduced to 1500, the
For 500 datasets, the accuracy, MSE, sensitivity, specificity, accuracy, MSE, sensitivity, specificity, and precision were

Volume 1 Issue 1 (2024) 52 https://doi.org/10.36922/aih.1746