Skip to main content

Table 1 The cost factors and data distribution in imbalanced retro-illumination images

From: Automatic diagnosis of imbalanced ophthalmic images using a cost-sensitive deep convolutional neural network

Actual class

Predicted class

Majority/negative

Minority/positive

Majority/negative (1970)

0

1

Minority/positive (735)

C min

0

  1. The numbers of positive and negative samples in the dataset were 735 and 1970, respectively. The cost factors for misclassifying positive and negative samples were C min and one respectively while the cost factor for correct classification was zero.