[rules-users] Newbie: Is this feasible w/ Drools?
Edson Tirelli
tirelli at post.com
Wed Feb 28 19:17:25 EST 2007
Alexander,
Seems a really interesting project indeed!
Trying to answer:
1. I'm not sure what you mean by "rule hierarchy". Drools does not
support rule "inheritence" if that is what you mean by hierarchy... can
you please ellaborate?
2. Dynamically adding rules to a rulebase will not force a full
reevaluation. Only the new rules are evaluated and if they eventually
have Patterns that may be shared, they are not even evaluated, as they
already were.
The problem I think you will face is that Drools does not support
uncertanty reasoning. You will need to code uncertanty by hand if you
use drools. I suggest you a POC to check how it goes.
[]s
Edson
Alexander Richter wrote:
> Hello,
> I work in a bioinformatics group and we're interested in using Drools
> for setting up an assertion engine for biological processes.
> We have over 300 genomes ( == sets of facts), each with upwards of
> 10K genes/features with annotations ( == facts). We want to be able
> to predict whether various biological processes exist, based on
> presence of the annotated genes that are the components of the
> process. Asserted processes are not black and white; we may have
> various levels of certainty, based on the percentage of component
> genes found (because we can't be certain that not finding a certain
> annotation means the the gene isn't in the genome -- it could mean
> that we just didn't find it). In addition, processes tend to be
> composed of sub-processes. The assertions then get used to make
> additional annotations on the features, which we then use to help us
> look for missing components. Therefore, we need to be able to display
> the decisions that led to a specific assertion, as well as the
> features that went into the component assertions.
> In addition, we regularly add new processes, as we get new ways to
> annotate genes, or read about newly worked-out processes. We really
> don't want to have to run all rules against all the genomes every
> time we add a new rule, but just add that rule.
>
> We would like to define the actual rules for the assertions in
> Drools, but don't know how we would do certain parts.
> 1) Can one walk through the rule hierarchy from a given rule and
> tease out the hierarchy of rules and the facts that went into that rule?
> 2) Is it possible to add a rule, or set of interacting rules, and
> have them incrementally evaluated, instead of forcing full
> evaluation? Alternatively, can the rulesets be individually defined
> and run in such a way that they interact?
>
>
> Thanks,
> Alexander Richter
> _______________________________________________
> rules-users mailing list
> rules-users at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users
>
--
Edson Tirelli
Software Engineer - JBoss Rules Core Developer
Office: +55 11 3124-6000
Mobile: +55 11 9218-4151
JBoss, a division of Red Hat @ www.jboss.com
More information about the rules-users
mailing list