Cancer Detection using Support Vector Machines Trained with Linear Kernels

In this article support vector machines are used for determining is cancer is present in lung, liver and cervix tissue using multiple kernels. The results indicate that linear kernel in this regard seems to be a better approach than using polynomial or Gaussian kernels. It was also found that using support vector machines trained with a linear kernel seems to also produce more accurate results than using a backpropagation neural network with 10 neurons. The accuracy of classification decreases when methylation in blood samples is analyzed, rather than direct tissue samples, to determining the presence of cancer.