Decision activity is a central part of business process. It helps the end-users to make better decisions by providing them information about previous decisions. The decision is to categorize a work-item into a pre-defined set of categories. It references a pre-defined set of simple variable types adn suggests the most likely categorization.
The exp file is in Avio-OBPM Training\Decision Activity\
create instance variables for decision
The attributes displayed in Decision Activity must all be simple data types.
initialize each instance variable except “creditRisk” in Log Receipt Activity.
currentAssets = creditRequest.currentAssets
isCurrentCustomer = creditRequest.isCurrentCustomer
requestedAmount = creditRequest.requestedAmount
residence = creditRequest.residence
Add decision activity and double click and setup like this
Add a global activity and setup its Main Task
and put in this code (java)
String[String] decisionFeatures;
decisionFeatures = {"currentAssets" : "DOUBLE",
"isCurrentCustomer" : "BOOLEAN",
"residence" : "STRING",
"requestedAmount" : "DOUBLE"};
String[] decisionClasses = {"Cancel","Low", "Medium", "High"};
DecisionProblem dp = new DecisionProblem(features: decisionFeatures, classes: decisionClasses);
dp.forceTraining(maxRecords : 10000);
To test it
create some “Credit Request”, categorize two as Low Risk, two as Medium Risk and tow of them as High Risk.
“Assess Credit Risk” in workspace
Run “Force Decision Activity Training”
Create more “Credit Request” and assess their credit reisk, the engine will now make recommendations on the likelihood that a request belongs in a given category. As you make more decisions and force the Decision Activity training, the recommendations will become more accurate.
I don’t really know the algorithm behind the scene, but the testing result is poor when I created around 10 requests…
No comments:
Post a Comment