DMZ Fuzzy Reasoning System Logo
732 369-6544

When you don't have enough data to mine

The DMZ Fuzzy Reasoning System allows predictive classification when insufficient data exists for mining. Its runtime can be part of a web application, a web service, client server application or a standalone application on PCs or mobile devices.

At the system's core is an expert system with fuzzy logic rules. This allows the domain experts to build their system with approximate inputs; the type of data that is usually available in real-world applications.

A key feature of the system is its authoring tool that allows domain experts to refine their rules one test case at a time, rather than guessing at global rules and how they may interact. When defining input data, the domain expert can specify the expected distribution of input data. The system then generates thousands of test cases, runs them through the runtime rules engine, and data mines the results. The domain expert can drill down to find individual test cases of interest. A full view of the rule execution for each test case allows identifying changes to the rule database. This process is performed until satisfactory results are obtained.

The system has facilities for domain experts to document their rules, provide long and short messages to users, and to attach recommendations for user viewing to the rules. Since an intermediate rule may trigger a recommendation, a facility is available to allow a forward chaining fuzzy rule set to refine the recommendation based on the execution of an intermediate rule.

Currently in tests with to provide advice to college applicants on how to complete their application, along with advice on things they can do to improve their chance of admission. This is an excellent example of a possible use of the fuzzy reasoning system; sufficient information isn't available for data mining, however considerable domain expertise exists without precise values around the rules. A prototype targeted to MBA students is available on their site and can provide insight into the application of the DMZ Fuzzy Reasoning System.