Linear and Non-Linear Classifiers for Clinical Risk Factor Analysis of Cancer Patients.

Cancer is a common disease condition, which is growing rapidly in the modern society. Different features such as tumor size, age, number of genomic alterations and non-synonymous mutations etc. have been considered in diagnosing and treating cancer patients. Linear and non-linear models can be utilized in analyzing the risk factors of cancer patients. Regression analysis is one particular machine learning technique, which could be perfectly utilized to identify and characterize relationships between multiple risk factors of cancer. As the insights of relationships among the clinical risk factors of patients can be useful in diagnosis and treatments of cancer, our main objective is to identify the relationships among their clinical risk factors using simple and multiple regression analysis as the machine learning techniques and find the appropriateness of the results with the previous experimentally proved results. A significant (p-value u003e0.05) relationship could be observed among the overall survival months and the age of merged cohort of Lower Grade Glioma and Glioblastoma brain tumor patients. Further, the relationship between the percent aneuploidy and the age of these brain cancer patients is observed as significant. Moreover, a significant relationship can be seen between the overall survival months and the age of Acute Myeloid Leukemia patients. Further, we analyzed the brain cancer dataset using regression trees as it can provide a solution model using overall survival months and percent aneuploidy over age and found a non-linear solution.