Online Program

SVM Model for Blood Cell Classification Using Interpretable Features Outperforms CNN-Based Approaches (308217)

Presentation

*William Franz Lamberti, George Mason University 

Keywords: Shape Classification, Blood Cell, Machine Learning Applications

This paper presents a competitive solution for blood cell classification using support vector machines with a polynomial kernel with interpretable features. This approach was able to achieve an overall classification rate of about 98\% and outperformed convolutional neural network based approaches by about 5\%. Furthermore, by using variables which have a clear meaning, we can create a model which is far more interpretable than its convolution neural network counterparts. The paper and code is available at https://github.com/billyl320/bccd_svm.