Evolution of the Probability of an MPM (Multiple Primary Malignancy) after a First Colon Tumor

Starting from a database containing over 30-years observations on multiple primary malignacies (SEER Program), the most probable tumor following a first colon cancer is determined, based on premises such as sex and age at first diagnosis of the patient. Decision trees and randomized algorithms are used in order to extract significant data from the whole database. Moreover, tree structure varies during the years, hence a dynamic decision tree is obtained by connecting decision trees computed on 5-years patient cohorts. The resulting probability for each district is presented by using a linguistic fuzzy interpolator. Thus, a linguistic tool for second tumor forecasting is obtained.