Probability Calibration with Fuzzy Set Theory to Improve Early Cancer Detection
Oksana Filimonova, A. G. Ovsyannikov, Natalya Victorovna Biryukova · Doklady Mathematics · 2023
Abstract Cancer is the leading cause of death before the age of 70 years. An important step for reducing the cancer mortality can be its early detection. To improve the early diagnosis of cancer, we propose a novel probability calibration method based on the fuzzy set theory. Our approach was tested on the detection of female breast cancer and lung cancer. These are complicated by a small data set for the first case and by highly imbalanced data for the second case. In both cases, our probability calibration method improved the Log Loss metric (the best result was improved by 48.86%), the Brier score (the best result was improved by 13.24%), and the Precision-Recall metric (the best result was improved by 13.94%). The application field of our algorithm can be extended to any progressive diseases and events without a clearly defined boundary.