Page 76 - Big Data Analytics for Intelligent Healthcare Management

4.6 RESULT AND ANALYSIS 69

Validation accuracy (%)

Xception+SVM
IRV2+KNN
IRV2+LR
IV3+SVM

RN50+KNN
RN50+LR
84 86 88 90 92 94 96
FIG. 4.8
Validation accuracy graph for 100 .

Validation accuracy (%)

Xception+SVM

IRV2+KNN
IRV2+LR
IV3+SVM
RN50+KNN
RN50+LR
84 86 88 90 92 94 96
FIG. 4.9
Validation accuracy graph for 200 .

4.6.2.4 Validation accuracy of 400×
Interpretation: On the 400 data, most of the combinations of feature extractors and classifiers gave a
validation accuracy above 86% and the ResNet50 and LR classifier gave the best cross validation score
of 92.03% (Fig. 4.10).

4.6.2.5 Best validation accuracy
Table 4.6 summarizes the best validation accuracy achieved. It is noticeable that for 40 , 100 , and
400 , the ResNet-50 with LR classifier performed better than any others.

4.6.2.6 Performance on the test set
To evaluate the performance of the combinations of feature extractors and classifiers, some parameters
are described below with the help of a sample confusion matrix. In this work, the positive class is
Malignant, which means cancer is present and the negative class is Benign, which means cancer
is not present.