Machine learning based sensitivity analysis for the applications in the prediction and detection of cancer disease

Abstract

Machine learning is used in almost all the medical fields by the diagnostics and doctors especially in predicting and detecting the risk of cancer. This growing trend of machine learning utilization in this approach enables the researchers to survey on the various types and approaches of machine learning or deep learning methods. Many methods are noted including an increased dependence on protein biomarkers and micro array data, increasing application leads to various types of cancer and instead of depending on an older Artificial Neural Network (ANN) methods, a newer trend of more interpretable machine learning methods are used. From the recent studies in the field, it is observed that machine learning or deep learning methods can be used appropriately in the range (20-30%) to improve the accuracy of prediction, development and progression, recurrence and mortality. In this paper, it is proved that unsupervised learning techniques could be used for predicting and detecting cancer tissues and appropriate analysis would be done on the data. The major merits of the proposed method over the existing cancer detection methods is the possibility of applying data from different types of cancer which describes the feature automatically and it helps to enhance the prediction and detection capabilities very specifically.