Combining Markov Models and Association Analysis for Disease Prediction

An approach for disease prediction that combines clustering, Markov models and association analysis techniques is proposed. Patient medical records are clustered and a Markov model for each cluster is generated to perform prediction of illnesses a patient could likely be affected in the future. However, when the probability of the most likely state in the Markov models is not sufficiently high, it resorts to sequential association analysis, by considering the items induced by high confidence rules generated by recurring sequential disease patterns. Experimental results show that the combination of different models enhances predictive accuracy and is a feasible way to diagnose diseases.