Knowledge
Builder for Capturing, Maintaining,
Deploying Business Rules in
eBusiness Systems
What are Business Rules?
"Business Rules" is the name used to describe the knowledge
underpinning an organization, allowing it to perform its functions
and potentially giving it its competitive advantage. It covers a
wide spectrum of knowledge, skills, know-how and expertise which
define an organization. "Business Rules" represents the
most important corporate asset and covers the following organizational
knowledge:
- Documented procedures, methods & policies
- Compliance with external rules, regulations and legislation
(e.g Tax and environmental regulations)
- Employee's know-how and expertise relating to customers, products,
services, resources, processes, operations and Risks.
The Capture and automation of business knowledge can deliver significant
competitive advantages to an organization which include:
- Preserving the expertise of key employees
- Applying business knowledge widely, accurately, consistently
& rapidly
- A key enabling technology for ecommerce; selling products and
services and supporting clients and partners over the World Wide
Web
- The ability to respond fast to business changes made possible
by the ease of maintenance of captured knowledge.
Technologies for Capturing & Automating Business Rules
The technology for capturing and delivering business rules has,
over the last 20 years, been given many "labels" including
"Rule Based Systems", "Knowledge Based Systems"
and "Expert Systems". These technologies share one basic
design concept that of separating the business rules from the rest
of the application running these rules. The business rules are maintained
in a separate file which is run by a generic software component
called the "Inference Engine". The Inference Engine is
designed to search for decisions based on facts (input data) it
receives and by consulting the rules in the rules base file. The
rules usually have the IF THEN format as in the example below:
IF Grade is Director
THEN decision is Pass expense claim
IF Grade is Senior Manager
AND Hotel class is A
THEN decision is Reject expense claim
IF Grade is Senior Manager
AND Hotel class is B
AND Department is Accounts
THEN decision is Pass expense claim
The main advantage of separating the rules from the inference engine
is the ease of development and maintenance of the business rules.
With user friendly rules maintenance tools, the business rules can
be developed, reviewed and maintained by the experts and decision
makers in a particular area of expertise (e.g risk underwriters
or process maintenance engineers). This ease of rules maintenance
enables an organization to respond fast to business changes and
is a key enabling technology for ecommerce.
Rules based technologies can suffer from two limitations; the lack
of effective tools for capturing knowledge from domain experts and
the limited deployment options for the Inference Engine (integration
& platforms). The XpertRule Knowledge Builder technology has
a number of unique features which overcomes these limitations:
- An enterprise knowledge development environment which is multi-user,
project based and with a graphical knowledge object explorer
- Extensive support for the capture & maintenance of rules
featuring graphical knowledge representation / structuring, knowledge
acquisition features and support for knowledge discovery from
historic data
- Flexible, Scalable & efficient deployment and integration
featuring COM+ and Java rules server, XML data exchange and Thin
web client architecture
Graphical Knowledge Representation in Knowledge Builder
In most business rules systems, the knowledge is represented in
the following format:
Once the number of such rules exceeds
20 or so, it becomes very difficult to get an overall picture of
the knowledge. Furthermore, it is very difficult to assess if a
list of such rules is incomplete or if it contains conflicting rules.
XpertRule Knowledge Builder in contrast, represents knowledge as
set of decision trees as shown below:
Note that the above two decision trees contain the same knowledge
as the previous list of rules. However, decision trees are more
understandable, give an overall picture and ensures no knowledge
gaps or conflicting rules. Decision makers also find it easier to
use trees than rules to express / maintain their know-how.
Furthermore, Knowledge Builder assist the developer in maintaining
the knowledge structure by displaying the overall hierarchy of decision
trees and attributes in a "knowledge map" as illustrated
in the following map:
In addition
to decision trees, Knowledge Builder also allows experts to express
their know-how in tabular rules format called decision cases as
illustrated in the example below:
Finally, Decision Trees in XpertRule
Knowledge Builder can also contain "procedures" which
are scripting blocks (in VB script or XpertRule format) for calculations,
string manipulations and for calling external programs. Below is
an example of a tree with an embedded procedure:
Knowledge Acquisition from Experts in Knowledge Builder
People with expertise gain such expertise through years of experience
in applying their skills. Such people can find it difficult to articulate
their expertise as a set of rules or a decision tree. Knowledge
builder allows experts to express their know-how in one of two easy
to express knowledge representations:
- A table of Decision cases representing examples of how they make decisions
- A table of Exception cases from a common decision
Knowledge Builder features a unique tree induction algorithm which
automatically derives decision trees from decision or exception
cases as shown in the example below:
Knowledge Discovery from data
Historic data captured and archived by an organization can be
a source of new knowledge. Two types of knowledge can be discovered
from such data;
- Data relating to decision making by experts such as risk underwriters
or trouble shooters can be used to discover the rules these experts
use in their skilled decision making process.
- Data relating to the performance of various business processes
can be used to discover business performance rules which can be
used to improve these processes. For example discovering the profiles
of bad credit applicants or suspicious insurance claims.
The XpertRule Miner technology uses tree induction from data to
generate decision trees with outcomes that have probability figures
attached. Such discovered trees can be used as knowledge modules
within Knowledge Builder.
The example below shows a decision tree generated from historic
data relating to Accept/Reject decisions made by loans underwriters.
The data table shows the application data and the decision made.
The discovered decision tree reveals 7 leafs (rules or profiles)
each with a probability of being rejected: 
Deploying Business Rules Applications using Knowledge Builder
In addition to its advanced knowledge representation and acquisition
features, Knowledge Builder supports the deployment of knowledge
components and applications on Wintel and Linux / Unix platforms
with full integration with other applications.
For Windows 2000 / NT / XP operating system, Knowledge Builder
applications are run using a COM plus inference engine ( Rules
Server)
with full ODBC, DLL & COM connectivity. This Rules Server can
be used in two modes:
- Interactive Q&A Inference mode with a PC display Client
or a thin web client (browser)
- Batch Inference mode whereby the rules server receives data
and returns decisions using XML. In this mode the Rules Server
provides rules processing for other applications or for a transactions
server.
For deployment on Linux / Unix and other Java platforms, Knowledge
Builder can generate the business rules and the inference engine
as a Java Knowledge Class (source code) which can be compiled to
produce high performance Java Rules Server. The Generated Java Knowledge
Class can be wrapped as a Servlet or an EJB. As with the Windows
deployment, the Java Rules Server can be used in two modes:
- Interactive Q&A Inference mode with a thin web client (browser)
- Batch Inference mode whereby the rules server receives data
and returns decisions using XML. In this mode the Rules Server
provides rules processing for other applications or for a transactions
server
Applications of Knowledge Builder in eCommerce
Customer Need Analysis
In order to recommend the best products and services to potential
customers it is important to understand their needs. This gives
the customer the confidence that we understand his requirements
and will offer a suitable product / service. This task needs knowledge
of the customer requirements and how they relate to the Company's
products and services. Customer need analysis involves an interactive
session of guided Questions and Answers whereby the next question
to the customer is driven by his/her profile and previous answers.
In certain areas, such as financial services, the line of questioning
has to ensure compliance with regulation regarding what information
has to be given to customers.
Customer Need Analysis application can either be run over the internet
or run by mobile sales force on a laptop computer. A Customer Need
Analysis module would normally format the captured data in XML format
for passing to other business modules such as "Product Selections"
or "Risk underwriting".
Product Selection & Recommendation
Having captured customer needs through a simple form or a knowledge
driven sequence of forms, the next step is to recommend to the customer
the most suitable products and services matching his requirements.
This can involve recommending a single product/service or ranking
a number of products/services by suitability (matching the needs).
A product selection Rules Server is normally passed an XML string
with customer requirements and products features data and it populate
this XML string with a list of products ranked by suitability. The
Rules Server may also select any optional product features and values
(for example which type of a Card and the credit limit on a Current
Bank Account).
Credit Risk Underwriting
The automation of the processing of applications for credit, loans
or mortgages requires the automation of credit risk assessment.
The knowledge for risk assessment can be captured from underwriters
and / or from historic data on applications and their subsequent
credit performance. Once captured, this knowledge can be deployed
as a Rules Server receiving XML applications data and returning
risk decisions. Note that credit risk decisions can be Accept, Reject
or Refer to manual underwriting. Risk decisions can also include
a credit limit figure.
Insurance Underwriting
This covers include Life, Motor and other insurance underwriting
application. Application (proposal) data in captured using XML forms
and is passed to the Rules Server. The initial decision made by
Rules Server is whether the application can be underwritten by the
server or if it need referral for manual underwriting. The next
decisions is to derive the premium (multiplier) and any special
conditions.
Insurance Claims Processing
These applications decide if an insurance claim can be settled
or if it should be referred for manual processing. The knowledge
involves capturing common problem patterns and norms for various
types of claims.
Customer Support and Help desk
Many organizations start their ecommerce initiative with developing
front end systems for "e-selling". However, these organizations
soon realise that a very import part of ecommerce is e-support,
which is the ability to provide systems for helping customers with
solving their problems relating to products and services. Such e-support
systems need to capture the knowledge required to provide help,
trouble shooting, problem solving and support. The knowledge captured
in e-support system falls into two categories; diagnostic knowledge
to identify the problem and advisory knowledge to recommend a solution.
The decision cases, decision trees and tree induction features of
Knowledge Builder are ideal for representing diagnostic and advisory
knowledge.
Trouble-shooting and advisory systems can be used directly by the
customers or by call centre agents to increase their problem solving
skills.
|
