Academia.eduAcademia.edu

Figure 20 – uploaded by Saba Bashir

See full PDFdownloadDownload figure
his article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DC 10.1109/ACCESS.2021.3110604, IEEE Access

Table 20 his article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DC 10.1109/ACCESS.2021.3110604, IEEE Access

Related Figures (23)

AN OVERVIEW OF STATE OF THE ART APPROACHES FOR CARDIOVASCULAR DISEASES PREDICTION TABLE |
AN OVERVIEW OF STATE OF THE ART APPROACHES FOR CARDIOVASCULAR DISEASES PREDICTION TABLE |
FIGURE 1: Detailed Architecture and Flow of Proposed Framewor!
FIGURE 1: Detailed Architecture and Flow of Proposed Framewor!
CLEVELAND DATASET ATTRIBUTES NAME AND DESCRIPTION
CLEVELAND DATASET ATTRIBUTES NAME AND DESCRIPTION
Starting with the proposed approach, individual classifiers are evaluated on all dataset. In Table 4, we present the performance of individual classifiers for Cleveland data set. The results show that Support Vector Machine, Naive Bayes and AutoMLP have better accuracy, sensitivity, specificity and F-measure as compared to other classifiers. The top-3 classifiers also have low classification error, absolute error and relative error. This helps us to construct Ensemble 1 (SVM, NB and AutoMLP). In Table 5, we have evaluated the performances of different individual classifiers for Hungarian dataset. The results show that Support Vector Machine, Naive Bayes and Neural Network have better accuracy, sensitivity, specificity and F-measure as compared to other techniques. Moreover, SVM, NB and NN also have low classification error, absolute error and relative error. This results in the construction of Ensemble 2 (NN, NB and SVM). A traditional confusion matrix is shown in Table 3.
Starting with the proposed approach, individual classifiers are evaluated on all dataset. In Table 4, we present the performance of individual classifiers for Cleveland data set. The results show that Support Vector Machine, Naive Bayes and AutoMLP have better accuracy, sensitivity, specificity and F-measure as compared to other classifiers. The top-3 classifiers also have low classification error, absolute error and relative error. This helps us to construct Ensemble 1 (SVM, NB and AutoMLP). In Table 5, we have evaluated the performances of different individual classifiers for Hungarian dataset. The results show that Support Vector Machine, Naive Bayes and Neural Network have better accuracy, sensitivity, specificity and F-measure as compared to other techniques. Moreover, SVM, NB and NN also have low classification error, absolute error and relative error. This results in the construction of Ensemble 2 (NN, NB and SVM). A traditional confusion matrix is shown in Table 3.
TABLE 5 PERFORMANCE COMPARISON OF INDIVIDUAL CLASSIFIERS FOR HUNGARIAN DATASET PERFORMANCE COMPARISON OF INDIVIDUAL CLASSIFIERS FOR CLEVELAND DATASET TABLE 4
TABLE 5 PERFORMANCE COMPARISON OF INDIVIDUAL CLASSIFIERS FOR HUNGARIAN DATASET PERFORMANCE COMPARISON OF INDIVIDUAL CLASSIFIERS FOR CLEVELAND DATASET TABLE 4
TABLE 6 PERFORMANCE COMPARISON OF INDIVIDUAL CLASSIFIERS FOR SWITZERLAND DATASET
TABLE 6 PERFORMANCE COMPARISON OF INDIVIDUAL CLASSIFIERS FOR SWITZERLAND DATASET
Table 8 presents accuracy, sensitivity, specificity and F- measure of individual classifiers averaged over the four heart disease benchmark datasets. The results show that SVM, NN and Auto MLP have better accuracy as compared to other classifiers. This results in another combination of classifiers for our Ensemble 5 (NN, SVM, AutoMLP). (NN, SVM, Perceptron). Similarly, in Table 7, we have evaluated the performances of different individual classifiers for Long Beach data set. The results show that Decision Tree, Naive Bayes and AutoMLP have better accuracy as compared to other techniques. Moreover, NB also has better sensitivity, specificity and accuracy. Additionally, DT, NB and AutoMLP also have low classification error. Furthermore, NB has low absolute and relative error as compared to other techniques. This results in the development of Ensemble 4 (DT, NB, AutoMLP).
Table 8 presents accuracy, sensitivity, specificity and F- measure of individual classifiers averaged over the four heart disease benchmark datasets. The results show that SVM, NN and Auto MLP have better accuracy as compared to other classifiers. This results in another combination of classifiers for our Ensemble 5 (NN, SVM, AutoMLP). (NN, SVM, Perceptron). Similarly, in Table 7, we have evaluated the performances of different individual classifiers for Long Beach data set. The results show that Decision Tree, Naive Bayes and AutoMLP have better accuracy as compared to other techniques. Moreover, NB also has better sensitivity, specificity and accuracy. Additionally, DT, NB and AutoMLP also have low classification error. Furthermore, NB has low absolute and relative error as compared to other techniques. This results in the development of Ensemble 4 (DT, NB, AutoMLP).
TABLE 8 AVERAGE ACCURACY, SENSITIVITY, SPECIFICITY AND F-MEASURE FOR FOUR DATASETS PERFORMANCE COMPARISON OF INDIVIDUAL CLASSIFIERS FOR LONG BEACH DATASET TABLE 7
TABLE 8 AVERAGE ACCURACY, SENSITIVITY, SPECIFICITY AND F-MEASURE FOR FOUR DATASETS PERFORMANCE COMPARISON OF INDIVIDUAL CLASSIFIERS FOR LONG BEACH DATASET TABLE 7
TABLE 9 COMPARISON OF ENSEMBLES SCHEMES FOR CLEVELAND DATASET
TABLE 9 COMPARISON OF ENSEMBLES SCHEMES FOR CLEVELAND DATASET
Finally, we have presented the average results of five ensemble vote schemes in terms of accuracy, sensitivity, specificity and F-Measure for four heart disease data set in in Table 13. The results show that Ensemble | has better accuracy for Cleveland and Hugarian dataset, Ensemble 3 has better accuracy for Switzerland data set whereas Ensemble 4 shows better results for Long Beach data set as compared to other techniques. We have combined the average results of five ensemble schemes in terms of accuracy, sensitivity, specificity and f-measure. The results show that Ensemble 1 (SVM, NB, AutoMLP) has better The ensembles are based on majority voting where the final class is selected which has highest number of votes from individual classifiers. It is also important to note that all results are computed using standard 10-fold cross validation.
Finally, we have presented the average results of five ensemble vote schemes in terms of accuracy, sensitivity, specificity and F-Measure for four heart disease data set in in Table 13. The results show that Ensemble | has better accuracy for Cleveland and Hugarian dataset, Ensemble 3 has better accuracy for Switzerland data set whereas Ensemble 4 shows better results for Long Beach data set as compared to other techniques. We have combined the average results of five ensemble schemes in terms of accuracy, sensitivity, specificity and f-measure. The results show that Ensemble 1 (SVM, NB, AutoMLP) has better The ensembles are based on majority voting where the final class is selected which has highest number of votes from individual classifiers. It is also important to note that all results are computed using standard 10-fold cross validation.
TABLE 11 COMPARISON OF ENSEMBLES SCHEMES FOR SWITZERLAND DATASET COMPARISON OF ENSEMBLES SCHEMES FOR HUNGARIAN DATABASE TABLE 10
TABLE 11 COMPARISON OF ENSEMBLES SCHEMES FOR SWITZERLAND DATASET COMPARISON OF ENSEMBLES SCHEMES FOR HUNGARIAN DATABASE TABLE 10
TABLE 12 COMPARISON OF ENSEMBLES SCHEMES FOR LONG BEACH DATASE COMPARISON OF ENSEMBLES SCHEMES FOR LONG BEACH DATASE
TABLE 12 COMPARISON OF ENSEMBLES SCHEMES FOR LONG BEACH DATASE COMPARISON OF ENSEMBLES SCHEMES FOR LONG BEACH DATASE
FIGURE 3: Average Accuracy, Sensitivity, Specificity and F-measure Comparison of Five Proposed Ensemble Vote Schemes average accuracy as compared to other ensembles schemes so this proposed ensemble can be used for heart disease diagnosis with high accuracy at real time. Figure 4 shows the graphical comparison of ensemble vote schemes (Ensemble 1, Ensemble 2, Ensemble 3, Ensemble 4 and Ensemble 5) for 4 heart disease data set in terms of accuracy, sensitivity, specificity and F- measure. It is clear from the Figure 3 that our proposed scheme (Ensemble 1) has higher accuracy as compared to other ensembles. It has achieved an average accuracy of 83%. ANOVA (Analysis of Variance) Statistics is used to perform significance testing of ensembles [44]. The proposed ensemble i.e ensemble | is compared with all the other ensembles and p value is calculated. The results of ANOVA statistics are shown in tables 15-18. The p value and f-stat in datasets indicate that the results are Statistically significant at 95% confidence interval. Therefore, the proposed ensemble (ensemblel) shows significant performance when compared with other ensemble classifiers. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: D 10.1109/ACCESS.2021.3110604, IEEE Access
FIGURE 3: Average Accuracy, Sensitivity, Specificity and F-measure Comparison of Five Proposed Ensemble Vote Schemes average accuracy as compared to other ensembles schemes so this proposed ensemble can be used for heart disease diagnosis with high accuracy at real time. Figure 4 shows the graphical comparison of ensemble vote schemes (Ensemble 1, Ensemble 2, Ensemble 3, Ensemble 4 and Ensemble 5) for 4 heart disease data set in terms of accuracy, sensitivity, specificity and F- measure. It is clear from the Figure 3 that our proposed scheme (Ensemble 1) has higher accuracy as compared to other ensembles. It has achieved an average accuracy of 83%. ANOVA (Analysis of Variance) Statistics is used to perform significance testing of ensembles [44]. The proposed ensemble i.e ensemble | is compared with all the other ensembles and p value is calculated. The results of ANOVA statistics are shown in tables 15-18. The p value and f-stat in datasets indicate that the results are Statistically significant at 95% confidence interval. Therefore, the proposed ensemble (ensemblel) shows significant performance when compared with other ensemble classifiers. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: D 10.1109/ACCESS.2021.3110604, IEEE Access
AVERAGE COMPARISON OF ENSEMBLE SCHEMES IN TERMS OF ACCURACY, SENSITIVITY, SPECIFICITY AND F-MEASURE
AVERAGE COMPARISON OF ENSEMBLE SCHEMES IN TERMS OF ACCURACY, SENSITIVITY, SPECIFICITY AND F-MEASURE
STATE OF THE ART COMPARISON OF ENSEMBLES FOR HEART DISEASE DATASETS This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
STATE OF THE ART COMPARISON OF ENSEMBLES FOR HEART DISEASE DATASETS This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: D 10.1109/ACCESS.2021.3110604, IEEE Access
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: D 10.1109/ACCESS.2021.3110604, IEEE Access
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
his article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DO 10.1109/ACCESS.2021.3110604, IEEE Access
his article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DO 10.1109/ACCESS.2021.3110604, IEEE Access
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ ASA aS ANOVA STATISTICS FOR LONG BEACH DATASET
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ ASA aS ANOVA STATISTICS FOR LONG BEACH DATASET
his article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DO 10.1109/ACCESS.2021.3110604, IEEE Access
his article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DO 10.1109/ACCESS.2021.3110604, IEEE Access
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved