Building Knowledge-Driven Decision Support System and Mining Data

“Knowledge should be shared. It only grows by sharing.”

This phrase finds its importance in today’s highly competitive and economically turbulent business world. Unless knowledge is shared among employees, it doesn’t take an organization anywhere. It’s important to share and manage it, in order to foster innovative thinking, develop and train employees and evolve into an ever-growing company.

Like it’s important to share knowledge within the organization, it’s equally important to determine what to share with whom. Not all details can be shared with everyone. This means that it is absolutely necessary to decide knowledge sharing rules and regulations, so that it can be used effectively and appropriately.

So, how do you think, knowledge is shared and distributed within an organization? What it takes to ensure its effective allocation and circulation? How do you automate the access and sharing of information?

Automating Knowledge Sharing

Implementing knowledge-driven decision support system is one of the best ways to capture, process and store and share knowledge among employees. The information can be easily accessed by the user to resolve a variety of problems, issues or concerns.

Before the development of knowledge-driven DSS, employees with high intellect had to perform knowledge-intensive tasks. An expert in a particular area would know how to approach a problem and go about it. Similarly, knowledge-based DSS asks relevant questions, offers suggestions and gives advice to solve a problem. The only difference is that it’s automated and speeds up the whole process.

What is a Knowledge-Driven DSS ?

A knowledge-driven DSS

• is a computer-based reasoning system

• that provides information, comprehension and suggestions to users

• to support them in decision-making.

It’s an integration of computerized business intelligence tools and technologies customized to the needs and requirements of an organization. So, the focus is on

• Identifying specific knowledge sharing and distribution needs of a company

• Setting objectives that need to be attained with a knowledge-driven DSS

• The selection of appropriate tools and technologies

• Understanding the nature of work and decision-making performed by its potential users

• Selecting data mining techniques

Key Terms and Concepts

A computer-based reasoning system is similar to any other type of decision support system when it comes to their architecture. But it turns into a knowledge-drive decision support system when artificial intelligence technologies, management expert systems, data mining capabilities and other communication mechanisms are integrated.

Before we dig deeper, let’s learn about few important terms and concepts used alongside knowledge-drive decision support system. This will help gain an in-depth understanding of such support systems.

1. Expertise: A knowledge-drive DSS comes with a specific problem-solving expertise. This expertise is based upon three components:

   • Knowledge in a particular domain and associated symptoms and signs

   • Understanding of the relationships between varied symptoms of a problems

   • Skills, ways or methods of solving the problem

2. Expert System: A computer system that imitates the decision making capability of a human expert is called an expert system or an artificial intelligence system. It is designed to solve problems by

   • Using if-then rules

   • Reasoning about knowledge

   • Drawing inferences from facts and rules

3. Knowledge Discovery and Data Mining: These are interrelated terms used for the process of extracting valuable knowledge and discovering patterns, in order to transform the knowledge into easily comprehendible structure for further use. Data mining is a buzzword but a misnomer. This is because data mining is a process of collection, storing and analysis of data and not finding patterns. Knowledge discovery goes through a series of steps:

   • Selection

   • Pre-processing

   • Transformation

   • Data mining

   • Interpretation

4. Development Environment: It’s the environment in which a decision support system is developed. It typically includes software for creating a DSS and knowledge base. The development environment may vary in size, depending upon production/development needs.

5. Domain Expert: A domain expert is a subject matter expert who has expertise/authority in a particular domain. A domain expert is an integral part of the team working on developing a decision support system.

6. Knowledge Engineer: A technical expert who integrates knowledge into a computer system when developing a decision support system, in order to solve complex problems that require human expertise.

7. Knowledge Acquisition: It is extraction/mining of knowledge from various sources, such as experts, databases and external programs.

8. Knowledge Base: It is the collection and storage of structured (facts, rules, regulations, characteristics, functions, procedures and relationships) and unstructured information that will be used by a DSS in decision making.

9. Interface Engine: It is a software system to simplify the conception and development of application interfaces between application systems. Typically, it’s a middleware application to transform, route and translate messages between various communication points.

10. Heuristic: It’s an approach to discovery and problem solving by employing practical methods. These methods may not be optimal but can help achieve immediate goals.

It’s important to be familiar with technical jargons that experts in this field use, in order to gain a deeper understanding of knowledge-driven DSS.

Characteristics of Knowledge-Driven Decision Support Systems

A knowledge-driven DSS is different from conventional systems in the way knowledge is extracted, processed and presented. The former attempts to emulate human reasoning while the latter responses to an even in a predefined manner. The main characteristics of knowledge-driven decision support systems are:

• These systems aid managers in solving complex problems.

• These systems allow users to interact with them during the process of decision making.

• The recommendations made by these systems are based on human knowledge.

• These systems use knowledge base that’s engineered keeping in mind the nature of problems they will solve.

• These systems aid in performing limited tasks.

• These systems use heuristic technique of problem solving.

Managing Knowledge-Driven Decision Support System Projects

Knowledge-driven decision support systems are expert systems that are developed when decision-making cannot be supported using traditional methods. A knowledge-driven DSS project goes through various stages and can be difficult to manage. It’s important to be committed to monitor the development of a knowledge-driven DSS.

Development Stages

• Domain identification (Choosing a subject matter)

• Conceptualization (idea formation, feasibility testing and commencement)

• Formalization (beginning with development officially)

• Implementation (completion and execution)

• Testing (fixing errors and modifications)

It’s important to monitor project development throughout very closely. It’s a collective effort of knowledge engineers, domain experts, DSS analysts, users and programmers. And a project manager keeps track of the scope, time, quality and budget, to ensure optimum allocation of resources and creation of a quality product. A project manager is a person responsible accomplishing the pre-decided objectives of a project.

Knowledge-Driven DSS Examples

Here are few examples of successful and popular knowledge-driven decision support systems:

• XCON (eXpert CONfigure): This expert system was built to decide the components required to build a complete operational system. Its job was to determine the spatial relationships among the components. The DSS configured VAX computers and was known as the largest rule-based knowledge-driven system for years.

• TAXADVISOR: As the name suggests, TAXADVISOR assisted attorneys by collecting client data and suggested actions that clients need to take to settle their financial profile. Its job was to aid attorneys in taxation and estate planning for clients with estates greater than $175,000.

• Life Insurance Selection Expert System: The expert system helped Meiji Mutual Life Insurance Company, one of the oldest insurance companies in Japan, deploys XpertRule to select the most suitable insurance product for an individual from among myriads of products.

Data Mining and Creating Knowledge

Before, data mining systems came into existence, businesses had statisticians studying data. They would look at the data, formulate a hypothesis and carry out a test to approve or disapprove it. But a data mining software doesn’t need to establish a hypothesis to be approved or disapproved. Rather it works in ‘discovery mode’ and looks for patterns.

Data Mining Models

There are two types of data mining models that can be deployed:

• Predictive Model: This data mining model predicts which prospects are likeliest to respond to a particular stimulus. It forecasts clear results based on patterns identified from known results. It takes into account the people who have already responded to a similar or same stimulus.

• Descriptive Model: This model describes patterns in existing data to create significant demographics subgroups, which can then be used for target marketing.

Data Mining Tools and Techniques

There are a large number of tools and techniques used to extract/mine data. Which technique is to be used depends on the type of data to be extracted.

1. Case-based Reasoning

   Case-based reasoning tools are used to determine the distance between or relationship among various components. A problem solved using this tool goes through 5 stages:

   • Presentation – the problem is described and entered into the system

   • Retrieval – the system matches it with the cases stored in the system

   • Adaptation – the system matches the retrieved closest-matching case and the problem to generate a solution

   • Validation - the solution then goes through a validity test and is justified if the user gives a positive feedback

   • Update – the valid solution is accepted and added to the case base in the system

2. Fuzzy Query and Analysis

   Fuzzy query and analysis is a data mining tool follows the mathematical concept for ‘fuzzy logics – the logic of uncertainty’ to determine results that are close to a particular criterion. Users can then pick one, depending upon his or her understanding.

3. Data Visualization

   As the same suggests, this helps analysts visualize complex relationships in multi-dimensional data. The benefit is that this tool graphically represents relationships among components from different perspectives. Statistical tools, such as regression, classification or cluster analysis are a part of this tool.

4. Genetic Algorithms

   Similar to linear programming models, genetic algorithms conduct random experiments by selecting the genes (variables whose values are to be identified) and their values at random to find the fitness function. The software will also combines and mutates genes to find optimized value.

Data Mining or Knowledge Extraction Process

Knowledge extraction is the process of identifying relationships between various components or symptoms. It’s about making the best use of data. Data mining or knowledge creation proceeds through a number of stages:

• Setting objectives

• Selecting data to be mined

• Run feature or cluster analysis to qualify data

• Selecting and applying an appropriate data mining tool

• Discover and apply knowledge to solve a problem

Data Mining Examples

Now you know what data mining is and how knowledge is extracted from data collection and analysis, let’s take a look at few data mining tools that companies are using.

• Siemens, a German multinational conglomerate and the largest engineering company in Europe, uses a decision support system that uses case-based reasoning tools, helping their technical support service staff answer the questions from current enquiry. The DSS uses results of previous enquiries as cases and retrieves close matching cases when a problem is entered into the system.

• ShopKo, a Wisconsin based chain of retail stores selling clothing, footwear, bedding, jewelry, beauty, houseware, etc uses a data mining project to find that the sale of film does not result in the sale of the camera. Rather it’s the sale of camera that affects the sale of films.

• Firstar Bank, now U.S. Bank, headquartered in Minneapolis, US, used data mining tools to determine the customers who were interested in knowing about their new products. They started doing the target emails, which increased the response rate.

• American Century Investments is a independent and privately controlled investment management firm, headquartered in Kansas City, United States. The firm makes use of data mining techniques to know who all from among their customers will be interested in buying their other products. They use the results to cross-sell their products.

Evaluating Development Packages

Whenever you decide to develop or buy a knowledge-driven decision support system software application, it’s important to consider following criteria:

• Development Features: Input rules, customizability, capabilities and maintenance

• Scalability: Ease of integration with other existing hardware and software, web technologies, operating systems

• Ease of Use and Installation: The ease with which end user will be able to work on it

• Security: Safety of data and company information

• Cost: Cost of technology, cost of development, maintenance cost

Knowledge-driven decision support systems help businesses solve problems and make decisions. However, a caution should be used when employing it. It doesn’t outsmart human intellect; rather it aids decision making.