Data processing: database and file management or data structures – Database design – Data structure types
Reexamination Certificate
1999-03-19
2003-05-06
Corrielus, Jean M. (Department: 2172)
Data processing: database and file management or data structures
Database design
Data structure types
C706S002000, C706S006000
Reexamination Certificate
active
06560592
ABSTRACT:
TECHNICAL FIELD OF THE INVENTION
The present invention generally relates to computer database storage systems and, more particularly, to a multi-model computer database storage system with integrated rule engine.
BACKGROUND OF THE INVENTION
Rules serve to model business logic. A rule is an atomic statement (a statement which can't be broken down to simpler statements) describing some aspect of the business's operation. Rules serve as guides for managers and employees for how they should act and what they should do in certain situations. These rules can be stored in the computer system. By doing so, the rules are easily accessible by all employees. As modern businesses must be very dynamic, a central repository also allows for easy rule modification and update. A smart database can use the rules in this repository to control its execution. These rules can cover all aspects of the business, from when to reorder supplies to the diagnosis of medical illnesses. Some rules have no applicability to a smart database (such as employee dress code), while others are much easier to use through the implementation of a smart database (like a 10000 rule set diagnosing respiratory illnesses). We will focus on rules which can be used to control the database; however, our system will allow for all types of rules.
As we are talking about a potentially huge number of rules (millions for a hospital which has numerous medical diagnosis systems as well as administrative rules), we must have some way to organize them all. We do this through the use of business objects. A business object encapsulates all the rules controlling a particular aspect of the business.
When a rule influences the database, we say that it “fires”. This is keeping with traditional expert system terminology. In order for a rule to fire, its condition must evaluate to true. The database may evaluate rules either implicitly or explicitly. While a rule, by itself, may be ambiguous as to when it is evaluated, its position in the set of business objects will dictate which way it is evaluated. In general, rules which are automatically evaluated control some aspects of the business whereas rules which are explicitly evaluated provide information for a query. Rules which provide information for a query are essentially the same as rules in expert systems. This aspect of our business rules engine is called a decision support system.
Decisions are often difficult to make. They require making seemingly ambiguous relations between the data. Fortunately, we can model this ambiguity in the form of fuzzy variables. For a certain piece of data, we may consider multiple values, and each value will have a certainty factor assigned to it. The mathematical operations relating these fuzzy variables are well defined.
Computers are not infallible because they are programmed by fallible creatures. Therefore, users will need to validate results after a query. A how and why mechanism will explain to the user why a particular decision was made. One of the advantages of the rules paradigm is that if the logic is flawed, the incorrect rule can be quickly and easily fixed. Sometimes the logic is not flawed, but just skewed. In this case, it usually means that two rules which are both contributing to a solution have incorrect weights. There will be a feedback mechanism such that users can identify if a decision was correct or incorrect. Based on that feedback, the relative weight of the rules will be adjusted. This is similar in mechanism to a neural network; however, a neural network is a “black-box” design. This mechanism will be overt and accessible to the rule set administrators.
Many times a decision must be made between various objects which must be evaluated on the basis of numerous properties. This problem usually falls under the classification of data mining. While these decisions can be modeled by rules, they often require an inordinate number of rules which is cumbersome at best. Instead, we can model these with a special type of rule, called a sorting rule. A sorting rule allows us to not only pick the best of class, but also a range, like the 5 best, or to even view all the data sorted by multiple values of varying importance.
While some values that we want to sort by, such as cost, have a well defined relation between values (a simple numeric greater than/less than), others, such as location, are not so well defined. Hence, we allow our users to create value maps, where they place the different values (such as city names) on a 2D (potentially 3D or maybe even 4D) grid to map the relative importance of one value towards another.
Some data, such as that often stored in blobs cannot be easily related even through maps. Therefore, we will also provide a neural network which can learn from the available data and assign according certain factors to various fields. We should provide neural nets which recognize the most common types of blobs, such as bodies of text, JPEGs or PNGs. Neural nets provide the additional advantage that they can deduce relations between fields of data that were not previously recognized.
What is interesting here is that we've gone from expert systems to neural nets, sometimes considered to be different ends of the AI spectrum.
FIG. 1
illustrates this relationship.
We start at rule-centric programming. Rules use fuzzy variables to aid in decision support. Rules are also used to define patterns when doing data mining. Pattern searching also relies on fuzzy variables to define relations between the elements. Data mining can also take advantage of neural networks.
Given the technologies that comprise a smart database, we must consider what makes the database actually “smart”. A smart database is one which uses the knowledge base available to it (in the form of business objects and learned neural data) to reduce the workload of the user. It will also potentially increase their accuracy and precision. It does this by enforcing relations between the data, automatically filling in fields and providing recommendations to the user to aid in the decision making process.
Everything that has been proposed here is possible with today's databases, except that the burden of providing a rules engine and/or neural network is put on the application programmer, who often has little experience with such components. Many rules can be handled with just triggers, except that many more do require a true rules engine, and we want a common interface for all rules. Further, most rules can be embedded in procedural code. However, that is where they are put today. But procedural code is difficult to maintain. Rules change often in today's business marketplace, hence IS departments want to make changes in real-time to the rules. They do not want to recompile an application or take down the database to link in the new code.
Further, IS departments do not want to add rules until they are tested and known to work properly. For this reason, a smart database must provide an environment for testing and debugging rules before they are put into effect. This environment will integrate seamlessly with the rule editor and feedback mechanism such that rules can be edited while a debugging session is taking place, and those changes will be reflected immediately in the session. As this environment is meant for testing and debugging, it will operate on a replicate of the data so that the original data will not be adversely changed.
We need to provide this dynamic environment because today's business environment is changing faster than anyone can keep up with. The rule-based environment relieves most of the burden off of IS departments, but we can aid IS departments even more with the addition of temporal information to rules. Temporal information allows rules to be in affect between certain times or after a certain date. This is particularly advantageous when a new policy takes effect sometime in the future, but we want to set up the database for the change now so that it automatically takes affect at the specified time.
The storage and use
Brugger S. Terry
Reid Jon D.
Corrielus Jean M.
Micro Data Base Systems, Inc.
Nguyen Tam
Woodard, Emhardt, Naughton Moriarty & McNett LLP
LandOfFree
Multi-model computer database storage system with integrated... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Multi-model computer database storage system with integrated..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Multi-model computer database storage system with integrated... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3034267