I recently got an update from the team at Sparkling Logic. I first blogged about them back in 2011. As noted previously, the product is built on a collaborative, social platform in which people work on multiple projects. Different roles are supported, allowing everyone to have their own todos and for teams to share these todos, assign them to each other etc. Social graphs show who has been working on what, allowing you to focus on groups of people working together in common areas. The default environment focuses not only on the rules- but on the cases or data in a selected set of real or sample data records. The “RedPen” mode allows a business user to interact with specific records or documents in a business-friendly form, adding and changing rule conditions directly on the documents to which the changes apply. Since I last spoke with them they have added a couple of new features:
First there is a heat map view. This takes the data-centric view and shows which fields in the data are used in rule conditions and which in rule actions with more rule usage showing as a stronger color. This allows you to see which fields really matter. Similarly, when viewing a single case in the form view, the rules that fired specifically for that case can also be reviewed easily using the form structure of the information as an outline. When viewing a business rule it can be projected on to the data being viewed to match the fields in the document being viewed to the conditions/actions of the rule. These features nicely extend the product’s focus on the data flowing the rules rather than just on the rules themselves making it easy to understand how the rules execute.
Second they have also added the ability to deploy as Java and .Net components using an inferencing engine based on work by Charles Forgy. This gives customers of the web-based development environment more options for deployment. In addition they have customers using the web-based engine that originally shipped with the product in production and other customers using the tooling/repository to manage rules while generating rules for execution in other commercial business rules products.
Finally they have introduced an add-on package for analytic techniques too. This allows what I call “data mining for rules.” A typical data set being manipulated by rules might involve hundreds of data items and thousands of records. What customers want is a set of rules that identify which records are, for instance, fraudulent. The new feature, called BluePen, allows a user to select an algorithm to identify the possible predictors of a target variable from a subset of the variables in the dataset. For instance, a dataset might include some known fraud records and the user would then select the field containing the “is fraud” flag as a target variable and select the fields that are not personally identifying as candidates. They use cloud processing to apply the selected algorithm to find the predictors. The system can use any algorithm available and plugged in through an open interface. Sparkling Logic has developed some of their own as well as integrated some developed by others.
Having found the predictive attributes, a second step uses these attributes to generate rules based on these predictors. The rules can then be reviewed and directly inserted into the rule management environment where they can be tested, simulated and compared with the results of other rules. Users can see the contribution of specific rules to overall metrics and more. The process is iterative and users can combine rules mined this way with those specified by experts or used historically.
You can get more information on Sparkling Logic here and they are one of the vendors in our Decision Management Systems Platform Technology report.