2025
Chapter
The paper examines a hybrid method of reasoning by analogy. This method combines case retrieval based on the domain knowledge with the use of the classical k-nearest neighbors (k-NN) data analysis algorithm. Domain-specific databases and knowledge bases used in this case are formed according to their ontologies, which explicitly determine their structure, restrictions on possible content and interpretation rules. Consequently, both case descriptions and derived solutions are represented in ontological terms and are understandable by domain specialists. The method is two-stage. At the first stage, formalized domain knowledge and reference databases are used to bring all case parameters to a common metric and evaluate their numerical equivalents. At the second stage, based on the evaluated equivalents, the similarity measure of cases is calculated using the k-NN algorithm. Such a hybrid approach enables flexible and accurate assessment of the similarity of situations in cases containing heterogeneous parameters. The proposed method has been successfully implemented in two applied domains. In medicine, where it was used as a basis for a medical decision support system for prescribing personalized drug treatment. In laser-based additive manufacturing using directed energy deposition technology, it supports process engineers on setting up operating modes of robotic equipment suitable for performing specific technological operations.