Expert system

An expert system (XPS ) is a computer program that can assist people in solving complex problems like an expert by wicking action recommendations from a knowledge base. Via so-called if-then relationships can be represented human knowledge ( relationships in the world) for computer course ( knowledge base). An expert system contains the functionality to create the knowledge base and improve ( knowledge acquisition component) to process ( problem solving component ) and the user to make understandable ( explanation component ). Expert systems are a part of the field of artificial intelligence. Examples are support systems for medical diagnosis or for analysis of scientific data. The initial work on appropriate software were made in the 1960s. Since the 1980s, expert systems are also used commercially.

History of development

The advent of expert systems has been accompanied by the failure of another research objective of artificial intelligence, which is often referred Tagged with General Problem Solver. Had they initially sought to use general problem-solving approaches to arrive at a system that would generate regardless of the problem area solutions, we soon found out that such a general did not realize was a problem solver and numerous questions achieved negligible results. Especially for problems in specific application domains a greater knowledge base for the generation of solutions was necessary. Expert systems are systems that are based on such, usually maintained by expert knowledge base. However, they reproduce not only the contents of the knowledge base, but are able to reach conclusions on the basis thereof to the other. The quality of an expert system can be measured by the extent to which the system is ever to conclusions in the situation and how flawlessly it goes about it.

Realization principle

Both the representation of knowledge as well as for drawing conclusions very different models are used:

• Case based systems are based on a case database which describe specific problems in context including one made ​​solution. The system attempts to locate a given case a comparable, similar as possible in case his case data base and transfer its solution to the current case. The concept of similarity of cases, just the key problem of such systems dar. Typical example of a case is a patient with his symptoms and the diagnostic test results. The desired solution would be a proper diagnosis here.
• Rule-based systems and business rule management systems (BRMS ) are not based on specific case descriptions, but to rules of the type " If A, then B". In contrast to cases such rules represent more general laws, from which conclusions are to be drawn for specific situations. Rules or business rules usually have to be updated directly by human experts in the system.
• Another approach that can be used in particular in classification problems is to systems which are self- learning processes in the position by means of decision trees. This is a form of inductive learning based on an example set. One example of such a series of attributes ( an object, such as a patient ), and their specific forms exist. The processing of such examples, the system undergoes a path ( see also search tree ): The individual attributes are nodes that emanate from them possible forms edges. Each of the system follows that edge that applies in the present example, continues this process attribute Attribute and finally reaches a terminal node (leaf). This finally gives the class, which is assigned to the object described. In the construction of decision trees is the goal to reach with small trees as possible to the best possible classification results. The difficulty here is the selection of the attributes.

Knowledge base

In an expert system or knowledge based system is the knowledge base (English knowledgebase ) the area of ​​the system that contains the knowledge in any form of representation. The knowledge base is complemented by an inference engine, which is a software that can be operated on the knowledge base.

Application

A need for expert system support exists wherever experts are missing or because of the problem complexity and the abundance of accumulated data material, the processing capacity of human experts is overwhelmed. The application effect of expert systems is the problem complexity and the level gap between an expert and the actual users proportionally. This difference in level is more serious, the more complex and diffuse the problem area. The latter is in turn stronger, more inhomogeneous the area relevant knowledge is structured and permeated the less the area of ​​formal, but is dominated by empirical knowledge.

Task classes and well-known expert systems

Typical task classes for expert systems are ( in brackets are the names of some unrealized expert systems ):

Disadvantages in the use

Expert systems can be counterproductive for the solution of a problem when users rely on them completely without intelligent supervision or no constant intelligent search for alternative solutions is operated. Because each expert system has only a limited amount of data, it only data from the immediate vicinity of the problem are usually fed. This creates the risk of missing important fundamental changes to only offer conservative solutions or explanations. The expert system can not the given parameters, the entire system into question (see closed world assumption ). Inventions, innovations, etc. require a creative combination of the problem with another - about once foreign - knowledge (eg that a candy bar slips unnoticed into a fuel tank, the expert system is not a gas station programmable value, which is why this case is not possible ).

When expert systems are automated, can in some application areas threatens to cause devastating effects, such as not intelligent assisted, automated military actions.

There is a widespread view that the Black Monday in 1987, contributed to or aggravated by the dynamics of many very similar reacting Computer Trader.