Predicting Health Status and Recovery Process of Cancer Patients

Patient Reported Outcomes (PRO) hold multidimensional information on the health status of cancer patients. This multidimensional information is retrieved through questionnaires on the quality of life of cancer patients and holds information on symptoms, harmful secondary effects and physical functioning from the date of the diagnosis on over the time of therapy. It would be beneficial to recognize patterns in clinical courses of former patients and use these for prediction in order to conduct better therapy strategies for new and present patients.

The goal of this thesis is to verify if prediction models on the health status of cancer patients can be developed based on their initial health status and therapy type. Based on anonymized medical cases and with the use of modern data mining techniques (traditional machine learning algorithms, time series analysis, case base reasoning) prediction models should be tested to recommend better individual therapies.