PC-Shell hybrid expert system

General characteristics

PC-Shell is the first Polish – commercially available – hybrid expert system shell . It has been developed as a result of earlier experiences gained while building PC-Expert expert system shell (1985-88) and domain expert system Diagnostician MC 14007 (1988). The PC-Shell received many awards and honors at international software fairs.

Applications

PC-Shell is domain-independent tool for building expert system applications. It can be used in any field: starting from banking and finances but on technical applications finishing. Typical areas of applications are:

  • intelligent decion support systems,
  • didactics (teaching) at universities and researches.

 It can be among others applied in such domains as:

  • financial and economic analysis,
  • analysis of loan applications in banks,
  • tax consulting,
  • thanks to the open architecture can be easily integrated in Management Information Systems, serving for example, for automatic analysis of economic indicators,
  • technique (technology), for example for the analysis of the measurement data and monitoring,
  • medicine, diagnostics and therapeutic recommendations.

 

 

Problem classes

Domain-independence of the PC-Shell system means that not the field but rather a class of problems determines the success ofthe application of this system. PC-Shell is particularly predisposed to solve problems of the following classes :

  • data analysis (interpretation),
  • classification,
  • diagnostics,
  • finance and banking,
  • investments,
  • marketing,
  • technique (technology),
  • dydaktics (teaching) of AI at universities,
Architecture

A it has been mentioned PC-Shell is hybrid system, it means it combines together differend methods of problem solving and knowledge representation. Among others PC-Shell has built-in, fully integrated, neural net simulator. Another important feature of the PC-Shell is the blackboard architecture, which allows you to divide a large knowledge base into smaller modules – thematicallyoriented, so-called. sources of knowledge. Additionally the knowledge sources can be heterogeneous in their nature.

Knowledge representation

With a hybrid architecture of the system PC-Shel,l are next to each other different methods of representing knowledge:

  • declative, rule-based (rules and facts),
  • triple : < object, attribute, value > OAV,
  • imperative form as a procedural programming language (possible part of knowledge base),
  • knowledge in the form of full texts,
  • knowledge distributed in neural nets,
  • opportunity to share the knowledge base into a number of knowledge sources (blackboard architecture).

The PC-Shell is equipped with its own knowledge representation language. This language thanks to the adopted solutions, including block structure, has the following features:

  • readability,
  • flexibility,
  • complete separation of expertise (declarative part) and control procedures,
  • ease of learning.
Explanation facilities

Extremely important and almost unique feature is the so-called expert systems. explanation. The PC-Shell provides a wide range of explanations encountered in today’s expert systems. These include the following types of explanations:

  • how explanations,
  • why explanations,
  • what is,
  • metaphor,
  • facts description.

Explanations HOW explain and report on how how the given problem was solved showing the way of logical derivation. The can be presented in the textual form or as a trees. Explanations WHY are justifying the appropriateness (reason) of questions put by the expert system. Explanations WHAT IS  clarify selected concepts being used in the knowledge base. In PC-Shell has been introduced an additional full text explanations, called METAPHOR, which are complementary to how explanations. FACTS DESCRIPTIONS shows the source and method of obtaining facts and the availability of further how explanations for them.

Inference/reasoning

PC-Shell provides three modes of logical inference: backward chaining, forward chaining and mixed reasoning. It should be emphasized the the inference engine is being using during reasoning the backtracking mechanism in similar way as in prolog systems. Matching mechanism is very flexible and doeas not require any declarations of variables using dynamic typing. As values variables, numbers as well as strings can be used.

Parameterization of knowledge bases

Important feature of the system is possibility of parameterization of knowledge bases. Thanks to that solution the dynamic (automatic) change of selected values (parameters) in knowledge base, without necessity of any “manual” changes in the source code of the base. A good example of the application parameterization mechanism can be a knowledge base, in which are testedsome values ​​of certain e.g. financial indicators in relation to certain thresholds or ranges. At the same time, some threshold values ​​can be variable, depending on the context. For example, you can evaluate some different financial ratios of companies belonging to different industries (branch). Similar problems arise in other areas, eg in the technique. PC-Shell he is facilitating it hind, letting also create categories of parameters.This innovative solution/mechanism gives many advantages to users, particularly knowledge engineers.

Database interfaces

PC-Shell is equipped with a standard interface for popular databases (such as dBase, Oracle, etc.) via ODBC. Knowledgeengineer has at his disposal a set of instructions for communicating with databases using SQL commandsThere is also the possibility of integration with other/external applications using built in in PC-Shell the DLL library.

PC-Shell is open

PC-Shell can be easily integrated with other applications using several different tools as needed. Among others, knowledge engineer is equipped with such techniques as: DDE, Ole Automation, set of commands for manipulating files and callin external applications and some others.

Manuals

The system is described in detail in the form of e-books in PDF format. You can easily download them from the homesite. Tofacilitate using the systems  there are provided plenty of examples (especially when combined with the demonstration of knowledge bases) to make easier to learn to use the system.

Capital Advisor – demo version

 

 

Invest Advisor – demo version

 

 

Polish tags: szkieletowy system ekspertowy, hybrydowy system ekspertowy, sztuczna inteligencja, sieć neuronowa