Page 82 - Big Data Analytics for Intelligent Healthcare Management

4.6 RESULT AND ANALYSIS 75

Table 4.12 Result of Inception V3 with LR, SVM, and K-NN on 100×

Feature Extractor Classifier Accuracy (%) Precision (%) Recall (%) F1-Score (%)
Inception V3 LR 90.89 92.10 94.70 93.38
SVM 90.65 90.67 96.11 93.31
K-NN 88.25 90.69 92.28 91.48

98
96

94
92

90
88
86

84
Logistic regression Support vector K-NN
Accuracy Precision Recall F1score

FIG. 4.17
Performance of InceptionV3 with three different classifiers for 100 .

Table 4.13 Result of Inception ResNet V2 with LR, SVM, and K-NN on 100×

Feature Extractor Classifier Accuracy (%) Precision (%) Recall (%) F1-Score (%)
Inception ResNet V2 LR 90.65 91.78 94.70 93.22
SVM 91.37 91.58 96.11 93.79
K-NN 91.85 94.31 93.64 93.97

Interpretation: With Inception ResNet V2, the Support Vector classifier gave the best recall value,
but K-NN gave the highest accuracy, precision, and f1score (Table 4.13, Fig. 4.18).
Interpretation: With Xception, the Support Vector classifier gave the highest recall but LR had the
highest accuracy, precision, and f1score (Table 4.14, Fig. 4.19).
4.6.3.4 Overall performance on 100×
Interpretation: The ResNet50 with LR gave the highest accuracy but Xception with the Support Vector
classifier had a higher recall value than others. On the other hand, with Inception ResNet V2, K-NN had
the highest precision (Fig. 4.20).