Predictive models, such as neural networks, support vector machines, cluster models, decision trees etc can be used in two ways. One is classification: you take some input data/facts and the model will estimate if this particular set has a given characteristic or not. E.g. given a set of genes, you might want to see if the individual is prone to suffer from a disease. You might be more interested in regression/prediction: given a set of input data/facts, the model will estimate a dependent variable. E.g. given a sequence of genes, the model will try to predict the next in line. Models can give probabilities, too. The model would have to be trained with an external data mining tool, but then could be imported by drools to provide a runtime execution environment.
Even then, before thinking of the particular implementation, I'd try to model the problem. Interviewing the domain experts and casting their words into an organized representation will not be easy, I'm afraid. We can share thoughts sometimes, if you want/can. Davide -- View this message in context: http://drools.46999.n3.nabble.com/How-to-write-a-Heuristic-Knowledge-in-Drools-Expert-tp4018012p4018027.html Sent from the Drools: User forum mailing list archive at Nabble.com. _______________________________________________ rules-users mailing list [email protected] https://lists.jboss.org/mailman/listinfo/rules-users
