4 Learning from Cases (reduction of the model)

Once the expert accepted the Case-Based Graph, a new rule-based knowledge base can be created, which contains only the informative attributes but gives the same evaluation for the cases as the ones used for the induction of rules. The reasoning uses rules but they are induced from the set of cases, thus this type of reasoning is called Case-Based Rule Reasoning. As the knowledge base is generated automatically by reducing an existing model, it is also called reduction. The attributes and the cases are already acquired, therefore there are no chapters discussing acquisition of attributes, acquisition of cases or knowledge import.

4.1 Benefits

The great benefit of the Case-Based Rule Reasoning is the reduced size, i.e. the significantly decreased number of the attributes. It enables the user to make a quick evaluation of new cases but attention is to be paid to possible loss of actuality. To avoid the use of outdated knowledge base, the original case-based knowledge base is to be maintained, constantly adding the new cases and regenerating the Case-Based Graph. If the conditions are changed, the Case-Based Graph will alter.

Tip: If having a rule-based knowledge base and, since it’s been in use, a number of cases are accumulated, perform the Case-Based Reasoning on the same knowledge base and then extract the rules generating a reduced knowledge base. Doing so, you are able to densify the knowledge described by the primer rule-based knowledge base.

4.2 Single-Level Hierarchy

The rule-based knowledge base generated by Doctus from the Case-Based Graph forms a single-level hierarchy of attributes. To distinguish this special kind of Rule-Based Graph from the ordinary ones, a different name is dedicated to it, which also appears on the third pane of Doctus (Figure G-23): Case-Based Rule Graph.

Apart from being single-levelled, this Rule-Based Graph can be used similarly to the ones built in ordinary rule-based knowledge bases, see chapter Original Decision – Hierarchy of the Attributes: the Rule-Based Graph (G-2.3).

Figure G-23: The Case-Based Rule Graph. (View Animation)

4.3 Missing or Indefinite Rules

If there were value ranges of some rules not covered or multiply covered by cases used for Case-Based Reasoning, in the rule set of the reduced knowledge base some rules may be missing or indefinite. (See Figure G-24)

Figure G-24: Indefinite Rules in Reduction.

The missing or indefinite rules may indicate impossible range or not well-defined attributes or values. Usually fine-tuning is needed to make these situations clear. The available operations of the rule set are the same then in rule-based knowledge bases, see chapter Original Decision – The Rules (G-2.6).

4.4 The Reasoning

Reasoning in case-based rule system works and looks the same as in rule-based systems (see chapter Original Decision – The Reasoning (G-2.7)), though without fine-tuning the evaluation of a new case(s) may be indefinite or none at all. In this second situation it is strongly recommended to repeat the Case-Based Reasoning with the new case(s) included.

4.5 Tacit Knowledge and Fine-Tuning

The missing or indefinite rules may be made definite by simply changing the outcomes of the rules manually. However, it is worth consideration, what caused these missing or indefinite rules? If the expert is sure, that it indicates an impossible range, the rule may remain missing or indefinite; if there is a new case(s) falling into that range, perhaps the conditions of the reasoning are changed, thus the refreshment of the Case-Based Reasoning should be considered.

If during the fine-tuning of the reduced knowledge base implied changes of attributes and/or values, these changes should be applied to the case-based knowledge base as well, and the Case-Based Reasoning should be repeated.

As the hierarchy of attributes in the reduced knowledge base is single-levelled, it can easily happen that there are more then 3-4 attributes, which makes handling of the rule set difficult. There is no obstacle to modify the graph into a multi-level one, using the same drag-and-drop technique as described in chapter Original Decision – Hierarchy of the Attributes: the Rule-Based Graph (G-2.3).

Tip: Before modifying the Case-Based Rule Graph, repeat the Case-Based reasoning choosing different benchmarks, for deeper understanding of the interdependencies of the attributes.

Fine tuning the Case-Based Rule Graph and using it as feedback to the original rule-based or case-based knowledge base the tacit knowledge is pulled to explicit domain.

4.6 Knowledge Export – Intelligent Portal

Doctus is capable of exporting knowledge bases in various forms of intelligent agents. (See Figure G-25) Some of these features are available in advanced mode only.

Figure G-25: Knowledge Export. (Take a Short Tour)

The exported knowledge may be:

• Static exported knowledge is some sort of “snapshot” of the knowledge base with no reasoning capability. Different types of graphics and tables can be generated, presenting graphs or details about the knowledge base. The report in rtf format, containing all tables is also available.
• Dynamic exported knowledge has capability of reasoning. Built-in templates are available to generate webserver (php, perl, jsp), client (html pages with enabled on-page reasoning, javascript) and desktop (c++, xls) applications.

Advanced: Build your specialized export templates based on the above listed predefined ones.

Using the Knowledge Export feature the exported knowledge base can be made available to various users, who will be able to use it for evaluation, though they will not be able to modify it. Some types of the exported knowledge are also appropriate to be placed into portals in forms of portlets. As reduction produces a rule-based knowledge base the exported versions of it are very similar to the deduction, the only difference is that here we usually have less fields to fill. (Figure G-26)

Figure G-26: The exported reduced model.