Modelling Technical Knowledge using the Bayesian Network and the GMDH Method

Abstract:

Knowledge is the competitive advantage of companies. Its source is experts who are embedded in the work environment and have technical knowledge - built mainly on the basis of procedures, documentation and professional experience. The rotation of experts causes a loss of company know-how. Therefore, it is important to formalise technical knowledge in order to further use it (e.g. to train new staff), to support decision-making and to draw conclusions from it. To this end, mathematical algorithms are used to organize, classify, infer and optimize it. On the basis of literature analysis and a case study, an approach has been modelled which allows to preserve expert knowledge, predict the level of knowledge and indicate the most important areas which build it by means of functions. The model consists of the following elements: (1.1) closed list of choices, which describing the knowledge area, (1.2) questionnaire template to acquiring knowledge, (1.3) formalized and acquired expert knowledge which is stored in the knowledge base, (2.1) the learned Bayesian Network, (2. 2) level of knowledge in the enterprise, resulting from the clustering of acquired technical knowledge, which is (3.1) the input data for GMDH Method, (3.2) the model of forecasting technical knowledge. Finally, this model is implemented in a real case study from the Research and Development department of a manufacturing company. Thanks to the use of the GMDH method, it is possible to define the most critical expert knowledge for the implementation of the new projects/orders and to build a knowledge level prediction model for a manufacturing company. The model based on the use of two algorithms - the Bayesian network and GMDH allows for a synergy effect, that is the acquisition of new knowledge in the company.