Page 55 - AIH-1-1

Artificial Intelligence in Health AI model for cardiovascular disease prediction

Figure 5. Performance analysis of artificial neural network algorithm on
the selected attributes.
Figure 7. Performance analysis of the K-means algorithm on the selected
attributes.

Table 3. GA selected attributes
S. No. Seven attributes selected with GA
1 Chest pain type
2 Resting electrocardiographic result
3 Maximum heart rate
4 Exercise-induced angina
5 Oldpeak
6 Number of major vessels colored
7 Thal
Abbreviation: GA: Genetic algorithm.
Figure 6. Performance analysis of artificial neural network-genetic
algorithms on the selected seven attributes. 72.6%, respectively, which were obtained. For 8 attributes,
the results for accuracy, MSE, sensitivity, specificity,
68.20% for accuracy, MSE, sensitivity, specificity, and and precision were 71.60%, 0.2840, 97.90%, 35.32%,
precision, respectively. From the analysis of the results, and 67.70%, respectively. However, when the number
the system performed better on 14 attributes. When of attributes was reduced from 14 to 10, the sensitivity
the cardiac disease attributes were reduced to 10, the decreased, and as the number of attributes reduced to 8,
classification accuracy, MSE, sensitivity, specificity, the sensitivity slightly increased (Figure 8). This shows that
and precision obtained were 58.90%, 0.4111, 63.90%, 8 selected attributes contributed to better classification
51.00%, and 67.50% respectively. For 8 attributes, the performance in this scenario. The MSE also increased as
same set of performance parameters was measured the number of attributes was reduced from 14, 10, and 8
at 58.90%, 0.4111, 63.90%, 51.00%, and 67.50%. The attributes.
performance of K-means classification, as the number of The performance analysis of the SVM classification
attributes was reduced from 14, 10, and 8, is illustrated technique with 14, 10, and 8 selected attributes was
in Figure 7. performed. The results obtained for the SVM classification
The performance evaluation of the system using the based on the accuracy, MSE, sensitivity, specificity, and
KNN classification technique with 14, 10, and 8 attributes precision were 84.10%, 0.1600, 93.80%, 67.70%, and
showed that the accuracy, MSE, sensitivity, specificity, and 81.80%, respectively. It was observed that when the
precision are 79.0%, 0.21, 95.80%, 54.60%, and 75.40%, attributes were reduced to 10, the performance accuracy,
respectively. Based on the result, the system performed MSE, sensitivity, specificity, and precision obtained were
better when 14 attributes were selected. It was also 77.80%, 0.222, 77.10%, 78.80%, and 84.10%, respectively.
observed that when the selected attributes were reduced to For 8 attributes, the results obtained for the same set
10, the classification accuracy, MSE, sensitivity, specificity, of performance parameters were 79.0%, 0.210, 93.6%,
and precision were 75.30%, 0.247, 93.80%, 48.50%, and 58.82%, and 75.86%, respectively. The MSE, specificity,

Volume 1 Issue 1 (2024) 49 https://doi.org/10.36922/aih.1746