Searching for Hidden Patterns That Affect the Overall Patient Survival with Data Mining
Abstract
This paper studies the issues that affect the survival time of patients with chronic lymphocytic leukemia, taking gender into account. The set of patients is divided into two disjoint subsets (classes) by the indicator of actual survival, whose value is less than the predicted value of overall survival. Nonlinear data transformations based on the calculation of the values of the class membership function for each attribute were used to detect hidden patterns in the analysis,. The threshold values between the classes on the numerical axis were determined, both by individual attributes and by generalized assessments of objects on defined sets of attributes. The threshold values were used to record logical patterns in the form of half-planes and display gender differences for predicting the survival of patients.