AN APPROACH TO THE CONTEXTUAL ANALYSIS OF TIME SERIES
Forecasting methods despite their conventions and limitations are the evolution of descriptive analytics mechanisms. Any model of the real-world objects works only under conditions of restrictions and agreements. The same conclusion can be made for the forecasting process, that it is not possible to forecast future state of the researched objects for 100%. However, building the most accurate forecast under the given conditions is the key. Modern data mining methods are based on a variety of models. However, such models can’t define the components of researched objects and processes except those contained in their models. The context allows using additional domain knowledge in describing the behavior of objects and processes in the form of qualitative assessments of their state. The same dataset in different domains will have various models and analysis results. The article deals with an approach to the domain context formation based on the ontology for analyzing time series of industrial processes indicators. The logical representation of the ontology based on the ALCHI(D) descriptive logic is also considered. The article describes as well experimental results confirming the correctness and effectiveness of the approach proposed.