An Intelligent Diagnosis of Anemia Using Tensor Flow

Abstract

Typically, hemoglobin levels are measured in the patient's blood to make the diagnosis. But as research develops, techniques other than assessing the conjunctival pallor may be used to determine hemoglobin levels. A framework consisting of decision tree, logistic regression comparison techniques, as well as the Histogram of Oriented Gradient (HOG), attempts to diagnose diabetes from connected photos. A consensus was reached on issues like room size 8x8, block size 8x8, bin number 15, L2Hys block normalization and entropy process, the best discriminator, random state is 10, minimum pollution reduction is 0.15, and the system can produce 82.5% of the test images, which had an image size of 256x128 based on the results of HOG and decision tree. As for the logistic regression is concerned, the best is blocking normalization and random state with unit size 16x16, block size 8x8, number of bins 11, L2Hys 30, system capacity, utilizing the Stochastic Gradient Descent (SGD) optimizer method. 92.5% of the best outcomes were produced in 24.20 seconds of calculation time.