Decision management
Blog: AuraQuantic Blog
Decision
management aims to design, build, and maintain systems that produce structured
or scheduled decisions.
Structured or
programmed decisions solve problems that arise frequently, and that already have
an established procedure to address them. In other words, we can generate a
structured decision to respond to a problem that is repeated with a certain
form of ductility and to define, analyze and forecast.
Decision
management leverages data generated by other systems such as business process
management (BPM) to improve operations, and enable fast, consistent, and
accurate decisions based on facts.
Initially, decision
management focused on insurance and loan underwriting, mortgage approval,
financial negotiation, logistics and public sector applications, but today it
is also being applied for fraud detection, risk management, cross selling,
supply chain management, etc.
WHEN SHOULD DECISION
MANAGEMENT BE USED?
We must make use
of decision management to solve complicated problems which are structured and
understood well enough to be solved by software. These tough decisions involve
one or more of the following:
- Multiple decision criteria and substantial calculations
- A large amount of data or various types of data
- Reusable decision logic
- Modified repeated decision logic
- Traceable and auditable decision making
- Compensation between competing objectives
- Multiple decision models acting together as sets
Trivial
decisions must be addressed through application logic, and decision management
is not necessary in this case. For this type of decision, AuraQuantic
incorporates different gateways that can redirect the process flow depending on
whether a series of conditions are met. These can allow one (convergent) or
several (divergent) outputs and apply the logical operators XOR, OR, and AND.
Most decision
management applies to repetitive operational decisions. For example, an
insurance company can develop a scoring model using machine learning, and a set
of rules to accept or reject parts based on certain parameters such as the
reliability of the client and the amount of the operation, and then apply this
automation to thousands of parts.
DECISION MANAGEMENT: AUTOMATION
OR SUPPORT
Each decision
requires a different decision management technique. In general, the structure
of an operational decision can be analyzed using the following model: observe,
guide, decide and act.
- In the observation phase, data is collected.
- The data is put into context and the implications considered in the orientation phase.
- The determination of what to do is done in the decision phase.
- And finally, in the action phase the response is executed.
Automating decisions
Many repeatable
decisions can be automated because the decision-making process is fully
structured and predictable.
For example, a
bank can develop an automated decision management model to accept or reject
credit card transactions and apply it to millions of transactions.
Decision management to support
decisions
In some cases,
it is not possible to structure all the logic involved in a decision and we use
decision management for support.
A software
decision service, based on rule processing or other prescriptive analysis, can
generate a proposal that is passed on to one person, who ultimately makes the
decision.
Returning to our
example of the insurance company, we could compile the score scales provided by
machine learning and the business rules to offer a first assessment to the
person in charge of validating a claim.
Decision support without
decision management
In many business
scenarios, analytics and BI are only used in the observation and orientation
phases.
This can be in
the form of BI reports, data visualization dashboards, business activity
monitoring (BAM) systems, or ML models for predictive purposes.
SELECT TECHNIQUES AND TOOLS
ACCORDING TO THE NATURE OF THE DECISIONS
Although there
are many techniques and tools to approach decision management, all of them make
use of rules to evaluate conditions and make decisions.
Rules processing
According to Wikipedia, ‘Business rules describe the operations, definitions
and constraints that apply to an organization’.
These types of
rules are appropriate for decisions based on deterministic factors found in
corporate policies, regulations, and also in the subjective judgment of
experienced entrepreneurs, or personal perception.
Currently some software
platforms such as AuraQuantic offer solutions to create and use in applications and include specific
engines to execute them.
Specifically,
AuraQuantic offers four types of business rules that can be applicable without
human intervention and therefore applicable to automated decision-making:
- Assignment: Provides the value of one variable based on the value of another.
- Textual: Your information is expressed through text and has unlimited storage and use.
- Inference: Returns the result as a consequence of combining a number of scores from different criteria.
- Calculation: Perform calculations of all kinds for any variable.
Machine learning
Being able to
anticipate market needs is every company’s dream. Traditionally, predictions
have been based on people’s analytical ability to draw conclusions from their
experiences.
However, thanks
to the capabilities of Machine Learning, this technology can now be used to
generate predictive models, which based on historical data are able to decide
which possibilities are more feasible.
However,
predictive analytics cannot perform decision management if it is not combined
with a rule.
For example, a
Machine Learning model can be used by an insurance company to quantify the
degree of reliability of a customer claim, but a rule such as “If the customer’s
degree of reliability is >9.5 and the amount of the claim is <$500, then it
can be approved ”.
Optimization
Optimization is
especially important in today’s business environment. And companies should use
the prediction of metrics and results to generate data that helps with decision
making.
Simulation
When we use
simulation software, the program introduces randomly generated input values
and returns results after multiple trials. This same concept can be used to
simulate the effect of a proposed decision service before it goes live and to
estimate alternative decision results.
Knowledge graphs
Knowledge is a
set made up of information, rules, experience, interpretations, relationships,
and connections in a context and in an organization, which constitute the basis
for action and decision-making
Knowledge representation
and reasoning is an area of
artificial intelligence that takes advantage of techniques such as knowledge
graphs or semantic networks to allow descriptive, diagnostic, predictive or
prescriptive analysis.
In other words, with a knowledge graph, the stored information is organized to make it accessible and useful.
In a decision
management model, the knowledge graphs are used to map the entities of the
decision model, and the path of the graphs is used to model the decision
process.
The post Decision management appeared first on AuraQuantic.
Leave a Comment
You must be logged in to post a comment.
