This document describes the Flex expert system toolkit, an expressive and flexible rule-based development system for building and delivering scalable and flexible expert systems and business rules applications.
Flex provides a comprehensive and versatile set of facilities for both programmers and non-programmers to construct reliable and maintainable applications.
Flex includes support for different types of rule-based inferencing.
The main two are forward-chaining production rules, ideal for data-driven reasoning and backward-chaining rules best suited for goal-based deduction.
The most common one in auditing is based on the idea of estimating frequencies using assumed statistical distributions.
The thinking is based on answering a question like this: Assuming the actual error rate is a particular number, X, and that the errors are randomly distributed through the population of items to be audited, how many items would we have to test and find error free to be Y% confident that the actual error rate is no worse than some other number, Z?
In addition, Flex is integrated with Flint which offers ways of handling inexact reasoning namely Fuzzy Logic, Bayesian Updating and Certainty Factors.
This means you can describe your business rules and processes, even when you do not have a complete functional description.
Alternatively, the rule-based component can be embedded within Java or C# or . In addition, Flex programs can be delivered straight on to the internet using Web Flex.
This function is based on real file type, dictionary keyword checks and regular expression checks, helping to protect your company from accidental or malicious data leaks while assisting with compliance efforts.
If you are involved with an expensive control check or audit that uses statistical sampling then this could be the most useful article you read this year.
Although Gaussian processes have a long history in the field of statistics, they seem to have been employed extensively only in niche areas.
With the advent of kernel machines in the machine learning community, models based on Gaussian processes have become commonplace for problems of regression (kriging) and classification as well as a host of more specialized applications.