Analysis on cancer subtype classification with deep reinforcement learning

Abstract

The word "cancer" denotes to a set of syndromes that can spread to various bodily areas and are brought on by abnormal cell proliferation. After cardiovascular illnesses, rendering to the World Health Organisation (WHO), cancer is the second largest cause of death in the world. To better understand molecular processes behind various cancer subdivisions, cancer categorization depends on gene expression information is essential. Conventional machine learning method have proven useful in this situation, but new approaches are needed for accurate and understandable categorization due to the difficulty and dimensionality of gene expression datasets. This article, we analyse various methods for the multiclass categorization of cancer subtypes that use deep structured reinforcement learning (DSRL). Our methodology addresses several significant issues in cancer subtype classification by combining the strength of deep neural networks with reinforcement learning. The analysis reveals the newly suggested model exceed the contemporary state-of-the-art classifiers, achieving a highest accuracy across all seven datasets, ranging from 55% to 100%, while also attaining the lowest loss, which varies between 0.02 and 0.11. This work offers a viable method for classifying cancer subtypes into many categories using gene expression data.