Machine learning for survival analysis: A case study on recurrence of prostate cancer

This paper deals with the problem of learning prognostic models from medical survival data, where the sole prediction of probability of event (and not its probability dependency an time) is of interest. To appropriately consider the follow-up time and censoring - both characteristic for survival data - we propose a weighting technique that lessens the impact of data from patients for which the event did not occur and have short follow-up times. A case study on prostate cancer recurrence shows that by incorporating this weighting technique the machine learning tools stand beside or even out perform modem statistical methods and may, by inducing symbolic recurrence models, provide further insight to relationships within the modeled data.