Cancer Literature Classification Methods Performance

Abstract

The literary classification system is the best solution to improve the data search process. In terms of the need, its goal is to compare the relevant biomedical papers and discover novel knowledge to identify potential research issues. This paper will present cancer literature classification performance by comparing three approaches, Naïve Bayes, Neural Network and Linear Classifier with SGD training. The propose approaches classify biomedical literature in five classes of cancer literature type namely, bone cancer, gastric cancer, kidney cancer, skin cancer and papillary thyroid cancer by using 9259 documents. General steps for building classification refer to the classification of scientific literature. The result shows that all algorithms successfully can be used to classify cancer literature. However, for the best performance, it is strongly recommended to use Naïve Bayes and Neural Network.