Page 57 - AIH-1-1

Artificial Intelligence in Health AI model for cardiovascular disease prediction

Table 4. A decision tree with 14 attributes selection
S. No. 14 Attributes
1 Age
2 Sex
3 Chest pain type
4 Resting blood pressure
5 Serum cholesterol
6 Fasting blood sugar
7 Resting electrocardiographic result
8 Maximum heart rate
9 Exercise-induced angina
10 Oldpeak Figure 11. Performance analysis of artificial neural network algorithms
11 Slope on datasets of different sizes.
12 Number of major vessels colored and 70.10%, respectively, for 2000 CVD datasets. When
13 Thal the dataset number was reduced to 1500, the accuracy,
14 Class MSE, sensitivity, specificity, and precision obtained were
42.00%, 0.5800, 48.20%, 38.30%, and 61.80%, respectively
Table 5. A decision tree with 10 selected attributes (Figure 12). For the 1000 CVD datasets, the accuracy,
MSE, sensitivity, specificity, and precision obtained were
S. No. Ten attributes 58.00%, 0.4200, 62.30%, 48.40%, and 72.90%, respectively.
1 Resting blood pressure The accuracy, MSE, sensitivity, specificity, and precision of
2 Serum cholesterol 500 CVD datasets were 54.00%, 0.4600, 55.00%, 50.00%,
3 Fasting blood sugar and 81.50%, respectively. When the CVD dataset was
4 Resting electrocardiographic result reduced from 2000 to 1500, the accuracy and sensitivity
5 Maximum heart rate decreased and increased at 1000 CVD datasets before
decreasing when the number of dataset equals to 500. The
6 Exercise-induced angina MSE increased when the dataset was reduced from 2000 to
7 Oldpeak 1500, then decreased at 1000 before increasing when the
8 Slope dataset was further reduced to 500, whereas the specificity
9 Number of major vessels colored and precision decreased as the dataset was reduced from
10 Thal 2000 to 1500, then increased as the dataset was further
reduced.

Table 6. A decision tree with 8 selected attributes The performance of the ANN-GA was analyzed on
the selected CVD datasets of 2000, 1500, 1000, and 500
S. No. Eight attributes (Figure 13). GA was used for the selection of a subset of
1 Chest pain type attributes. The seven best attributes that are more correlated
2 Resting blood pressure are responsible for cardiac disease. After training and testing
3 Serum cholesterol the ANN with the obtained attributes, the performance of
4 Fasting blood sugar 2000 CVD datasets in accuracy, MSE, sensitivity, specificity,
5 Max. heart rate and precision were 80.00%, 0.2000, 97.20%, 54.20%,
6 Exercise-induced angina and 76.10%, respectively. For 1500 CVD datasets, the
accuracy, MSE, sensitivity, specificity, and precision were
7 Oldpeak
77.80%, 0.2220, 95.70%, 59.10%, and 71.00%, respectively.
8 Number of major vessels colored For 1000 CVD datasets, accuracy, MES, sensitivity, and
precision of 73.30%, 0.2667, 90.00%, 40.00%, and 75.00%
After classification using the K-means technique, were obtained respectively, and for 500 CVD datasets, the
the performance of the selected datasets was evaluated same set of performance parameters was 73.30%, 0.2667,
in terms of accuracy, MSE, sensitivity, specificity, and 83.00%, 66.70%, and 62.500%, respectively. The accuracy
precision, measuring 59.50%, 0.4050, 64.10%, 51.40%, and sensitivity decreased steadily as the number of datasets

Volume 1 Issue 1 (2024) 51 https://doi.org/10.36922/aih.1746