Ch. 10
Building Knowledge-Driven DSS and Mining Data

We can identify a number of characteristics that are common to Knowledge-Driven DSS. First, this category of software aids managers in problem solving. Second, the systems use knowledge stored as rules, frames, or likelihood information. Third, people interact with a program when they are performing a task. Fourth, Knowledge-Driven DSS base recommendations on human knowledge and assist in performing very limited tasks. Fifth, Knowledge-Driven DSS and expert systems do NOT "think".

A Knowledge-Driven DSS differs from a more conventional Model-Driven DSS in the way knowledge is presented and processed. This difference exists because most expert systems attempt to simulate human reasoning processes. A Model-Driven DSS has a sequence of predefined instructions for responding to an event. In contrast, a Knowledge-Driven DSS based on expert system technologies attempts to reason about a response to an event using its knowledge base and logical rules for problem solving. Expert system technologies use representations of human knowledge. These representations are expressed in a special purpose language such as OPS5, PROLOG or Lisp. Expert systems can also perform standard numerical calculations or data retrieval. An expert system development environment uses heuristic methods to obtain a recommendation. A heuristic is an approximate method that identifies varying amounts of uncertainty in conclusions. A conventional Model-Driven DSS uses mathematical and statistical methods to obtain a more precise solution.

Figure 10.1 shows the components of a Knowledge-Driven DSS. The inference engine is the software that actually performs the reasoning function. In small systems, this is sometimes called the shell of the expert system, though the shell can be considered to be everything except the knowledge base itself. The inference engine is the software that uses the knowledge represented in the knowledge base to draw its conclusions. The design of the inference engine may limit the ways in which knowledge can be represented in the knowledge base so that certain shells are only suitable for particular types of applications.

In comparing Knowledge-Driven DSS and Model-Driven DSS, we should remember that: