Learn
Imagine software that can learn how your top experts make critical business decisions, and can then apply that knowledge reliably over millions of transactions; software that never tires, or loses concentration. A system that learns incrementally, and takes into account all of the available information – that's TIBCO® Patterns – Learn. It uniquely models human decision-making based on examples provided by your own business experts. As the machine "sees" more and more examples, it "learns" to make decisions as well as your best humans.
With TIBCO Patterns – Learn, there is no need to develop, refine and maintain complex rules. What's more, it continues to learn as you provide ongoing feedback. Because machine learning is quick and efficient, models can be custom tailored to the specific data-sets, markets, and business requirements of your application.
- Models the decision-making of your own domain experts
- Automatically creates models customized to your specific data, markets, and businesses
- Learns to solve problems automatically with no need to build, tweak or manage complex rules
- Improves incrementally as additional examples are provided
Implementing TIBCO Patterns – Learn starts with the definition of a decision. Do two or more records represent duplicates? Should they be linked? Do they represent the same entity? These, and any other data-driven "yes/no/maybe" question, can be modeled.
Typically, input records are generated by the TIBCO matching engine. However, this is not a requirement. Input records can be generated by other matching software, or any other process.
The model is automatically generated from sample decisions, for example whether two records represent a duplicate. The examples are paired records where experts have decided "Yes, these records represent a duplicate," or "No, these records are different." There is no need for humans to explain the reasons for the answer, or for programmers to infer underlying rules. The system draws its own inferences, and automatically picks up on subtle inter-field relationships that non-experts often miss. The software continuously builds its knowledge and adjusts as it learns to refine the accuracy of results. It can advise users when additional learning is needed, further enhancing its capabilities as more and more human insight is shared with the software.
