Real-Time Clinical Decision Support System with Data Stream Mining

Table 2

Defects of traditional implementations.

Algorithm	Defect
Traditional decision tree	Only can analyze static and finite data set. Cannot handle data stream
Bayesian network	Difficulty to get the probability knowledge for possible diagnosis and not being practical for large complex systems given multiple symptoms
Neural network	Training process consume so much time that users cannot use the systems effectively
Rules	It is difficult for experts to transfer their knowledge into distinct rules, and it needs many rules to make system effectively