Page 87 - Big Data Analytics for Intelligent Healthcare Management

80 CHAPTER 4 TRANSFER LEARNING AND SUPERVISED CLASSIFIER

Xception+KNN
Xception+SVM
Xception+LR
IRV2+KNN
IRV2+SVM
IRV2+LR Recall
IV3+KNN Precision
IV3+SVM Accuracy
IV3+LR
RN50+KNN
RN50_SVC
RN50_LR
82 84 86 88 90 92 94 96 98 100
FIG. 4.25
Test performance graph for 200 .

Interpretation: The ResNet50 with LR and Inception ResNet V2 with Support Vector classifier had
the highest accuracy but the Xception with Support Vector classifier had the highest recall value while
ResNet50 with LR had the highest precision (Fig. 4.25).

4.6.3.6 Test performance on 400×
Interpretation: With ResNet50, the Support Vector classifier had the maximum value for accuracy,
precision, recall, and f1score (Table 4.19, Fig. 4.26).
Interpretation: With InceptionV3, the Support Vector classifier had the maximum value for accu-
racy, precision, recall, and f1score (Table 4.20, Fig. 4.27).
Interpretation: With Inception ResNet V2, the Support Vector classifier had the highest accuracy,
precision, recall, and f1score (Table 4.21, Fig. 4.28).
Interpretation: With Xception, the LR had the maximum accuracy, precision, recall, and f1score
(Table 4.22, Fig. 4.29).

Table 4.19 Result of ResNet-50 with LR, SVM, and K-NN on 400×

Feature Extractor Classifier Accuracy (%) Precision (%) Recall (%) F1-Score (%)
ResNet-50 LR 91.48 92.94 94.80 93.86
SVM 92.86 93.75 96.00 94.86
K-NN 89.65 91.63 93.12 92.37