Data mining occurrences of infectious diseases with SNOMED CT

Abstract

Synonyms within SNOMED CT’s structure give meaning to the clinical terminology. The hypothesis in this thesis is that the number of synonyms of a disease within SNOMED CT can be used to predict the number of occurrences of an infectious disease reported on by the World Health Organization (WHO). Using simple Classification and Regression (CART), Bayes theory, and Best Fit trees, prediction algorithms are created based on the number of synonyms in infectious disease terms of SNOMED CT, the number of those diseases world-wide, the region of occurrence of the disease, and the year of occurrence of the disease. The results of experiments predict the number of occurrences of a disease correctly 67% of the time by using Simple Cart method; Bayes and Best Fit Trees each produce the correct number of occurrences 61% of the time.