|
Works "out ot box" for most matching problems Most Netrics customers find that their matching problem - even challenging ones - can be solved by the Netrics Matching Engine. The only "customization" required is to design the queries: to decide which fields should be used (individually or in combination), and the relative weights to assign to each field or querylet. Often, good results are obtained from the initial trial query. Careful optimization can take a few hours or days. The simplicity of the matching task is the direct result of the advanced, proprietary technology built into the Matching Engine. This patented, fourth generation matching algorithm was invented by Netrics' founder and uses bi-partite graph mathematics to model human perception. It's so advanced, it's simple. This is in stark contrast to conventional fuzzy matching technology which is based on complex rule-sets and probabilistic models. Creating a new rule set to deal with a new matching problem is a major programming and requirements definition task, often measured in man-months or even man-years. Even if the software comes with a pre-built rule-set, your problem will often not exactly match the assumptions of the original developers and will require modification. Such modifications are a major task in themselves, and fraught with the possibility of "breaking" the validity of the original model. For the most challenging applications, Netrics users can deploy the "Decision Engine". This adds multi-field logic to matching decisions, not unlike a rule set. However, there's one huge difference: the system learns the rules itself. You need only provide it with examples. There's no programming, no requirements gathering, and no manual translation of business rules to code. Using the Decision Engine, if new cases are discovered, you need only expand the training set. The system incrementally adds decisions about the special cases to what it already "knows". And it tells you if the training you've provided has converged on a reliable solution, or if more examples are needed. There's no guesswork about whether your model is valid. Whether you use the Matching Engine by itself, or in combination with the Decision Engine, you can afford to deploy Netrics technology wherever it's needed.
|
