Developing knowledge-based system on the basis of explicit knowledge (rules) is not sufficient to model the knowledge at a granular level. To cover the granularity of the knowledge in medical domain or any other domain, mining rules from the data would be required. That's why contemporary decision support system consider both expert-driven  and data-driven approaches to build the knowledge bases. In this way, data mining approach will support to make the hidden knowledge explicit for the domain experts.

BrHanu,you use data mining in situations were the explicit knowledge or rules to develop the required system is/are not handy. What we do  then is to use data to learn the model that would represent the knowledge we are interested in. In contrast, if we have deterministic heuristics to implement the system without machine learning or data mining, it would make sense we follow the deterministic approach since that would give us more accurate result.

Yes, I think so. It can expands the functionality of the knowledge-based system.

Developing knowledge-based system on the basis of explicit knowledge (rules) is not sufficient to model the knowledge at a granular level. To cover the granularity of the knowledge in medical domain or any other domain, mining rules from the data would be required. That's why contemporary decision support system consider both expert-driven  and data-driven approaches to build the knowledge bases. In this way, data mining approach will support to make the hidden knowledge explicit for the domain experts.

I agree with the colleagues, but I think that data mining can help now on small tasks. For instance, if a doctor is seeing an outpatient and may assign a ICD-10 code to the chart, you may present the top-10 codes that doctor used  before, or the top-10 used by doctors of the same specialty. Not glamorous, I know, but it qualifies.

In order to build a Knowledge Based System you need to extract knowledge. It may be from experts, from printed sources. But in many modern applications such explicit knowledge is not available. In that case data may be a good source. And to churn out knowledge from data data mining techniques will be helpful.

In my opinion non-availability includes all such constraints, copyright or other legal issues. Even linguistic knowledge or lack of it may lead to  non-availability for all practical purposes.

Dear Brhanu Desta ,

knowledge-based system uses explicit knowledges which could be facts or production rules as in expert system. It seems obvious if we need to use implicit knowledge which are result of datamining in the objective to perform knowledge based system, we do it. It depends on the applications.

Data mining can help you to extract hidden knowledge from data such as patients databases, and thus you can feed your knowledge base with extracted knowledge beside the knowledge you have from whatever books, experts or internet.... Data mining is done by applying the suitable machine learning techniques to the existent databases.

Are you targeting a specific area in your proposed system?

Thank you for your answer!

Yea my target is specific, I am trying to develop a knowledge based system for malaria diagnosis and management in my country Ethiopia and I want to use data mining as well.

So guys please help me as much as you can

There are many ways to build a knowledge based system. You may extract explicit knowledge that represents the domain expert's heuristic and encode it in form of production rules in the knowledge base. You may equally look towards a memory-based or case-based or instance based approach(which is a data mining strategy popularly referred to as K-nearest neighbour algorithm).Also, you may want to develop a predictive model using historic data. Each of these approaches depend on what you want to achieve and the amount of domain knowledge available to you. Predictive model can be used only when your output is expected to  be numerical, either discrete( in the case of a classification problem i.e. 1 OR 0) or continuous( in the case of a regression problem i.e. any real number). To the best of my knowledge, I think you would need a memory based or case-based approach where your outputs are prescriptions or recommendations to each instance or example in the historic record or data set, such that for any new case, recommendations made to similar cases should be used for the new case. But,  knowledge-based systems are easily modelled when the domain knowledge is explicit and can be represented in IF-THEN production rules... but despite that no system is void of some degree of uncertainty, as such methods like Bayesian reasoning or certainty factor can be used to take care of such uncertainties that may crop up in the system.

I don't know if this would be of helpful to you.............

Dear Chinedu Pascal Ezenkwu

I thank you! It was helpful for me and now I am approaching to the point.

Data mining algorithms are generate or produce so many rules that are confused to choose the best rule to make a decision and the rules have too far from facts so decision maker doesn’t believe that rule to make a decision in real world decision making process. So using data mining alone for knowledge based system is makes u'r KBS accuracy decrease and low acceptance , better to be include domain expert validation.