3. Results Here, the results and evaluation of the diagnosis of heart disease using GA based trained RFNN approach is
presented. The results were obtained using Intel i7 CPU PC with 16GB of RAM, Ubuntu OS and Java. Table 2
presents the evaluation performance of the GA based trained RFNN approach. The RMSE result of the training set
is 0.0357, testing set result is 0.0222 and overall result is 0.0337. The testing set (45 instances) results showed that
the GA based trained RFNN approach had a sensitivity of 100%, a specificity of 95.24%, a precision of 96%, and F-
score of 0.9796 and accuracy rate is 97.78%.The approach has an excellent rate of 100% for the patients without
heart disease that were found to have no heart disease in the testing set (20 instances). The overall (297 instances)
results had a sensitivity of 97.74%, a specificity of 95.734%, a precision of 94.89%, and F-score of 0.9626 and
accuracy rate is 96.63%.
Table 2.Results of the diagnosis of heart disease using GA based trained RFNN
TN FN TP FP Total RMSE Sensitivity
(%)
Specificity
(%)
Precision
(%)
F-Score PME
(%)
Accuracy
(%)
Training
set 137 3 106 6 252 0.0357 97.25
95.8
94.64 0.9593 3.57 96.43
Testing set 20 0 24 1 45 0.0222
100
95.24
96 0.9796 2.22 97.78
Overall
157 3 130 7 297 0.0337 97.74
95.73 94.89 0.9629 3.37 96.63
Samuel et al. (2017) partitioned the Cleveland heart disease dataset into three sub-sets for training (193
instances), validating (59 samples) and testing (45 instances). Samuel et al. (2017) compared their results with the
conventional ANN and seven previous works to denote that their approach had better accuracy. The testing set
performance comparison of the GA based trained RFNN approach and ANN-Fuzzy_AHP is given in Table 3. Based
on this comparison the GA based trained RFNN approach had better accuracy than ANN-Fuzzy_AHP method.
Table 3. Performance comparison of the testing set
Author
Method
TN FN
TP
FP
Total
Accuracy (%)
Samuel et al. (2017) ANN-Fuzzy_AHP
20
0
21
4
45
91.1
proposed
GA
based
trained
RFNN 20 0 24 1 45
97.78