Supervised link prediction in symptom networks with evolving case

• A link prediction method to identify the connections between symptoms.
• A large number of correlations are hinted by the topology of a symptom network.
• Plentiful latent information from which inferring future disease interactions.
• Medical histories of a set of similar patients are important for prediction.
• Good accuracy, prediction and recall values for the case-based link prediction.

Medical care can improve the life quality since a patient can modify his habits and lifestyle in order to prevent the occurrence of probable correlated future symptoms causing to a disease. In this paper, we predict the onset of future symptoms on the base of the current health status of patients. The problem of predicting the relations between symptoms (abnormal parameters in this paper) which can be shown as the reason of a disease in the future is a really difficult and, at the same time, an important task. For this purpose, the present paper first constructs a weighted medical data network considering the relations between abnormal parameters. Then, we propose a link prediction method to identify the connections between parameters, building the evolving structure of medical data network with respect to patients’ ages. To the best of our knowledge, this is the first attempt in predicting the connections between the results of laboratory tests. Experiments on a real network demonstrate that the proposed approach can reveal new abnormal parameter correlations accurately and perform well at capturing future disease risks.