ROLAP-ML

The concept of Relational On-Line Analytical Processing based on symbolic machine learning methods – ROLAP-ML.

Form of financing: targeted grant

The total cost of the research project is 498,325.00 PLN.
A targeted grant to cover eligible expenses in the amount of 398,660.00 PLN.

The project aims to apply machine learning methods, specifically data pattern detection techniques, to automate the process of discovering strong, user-hypothesis-driven relationships in databases while maintaining user interaction. Methods for automatic data summarization will be developed for various exploratory problems, including group description (classification and subgroup discovery), explaining changes in variable(s) values (regression, survival analysis, and reliability analysis), explaining differences between groups (contrast set mining), identifying exceptional situations (such as exception rule mining), and action planning problems. Research in this area has been conducted by the project team for many years. The development of new methods will focus on enabling their operation in distributed computing environments.

The project will involve experimental work, including the analysis of benchmark datasets (from Kaggle and UCI repositories) and three proof-of-concept analyses demonstrating the usefulness of the developed methods. The result will be a set of tools that, either automatically or interactively with the user, searches databases to discover interesting relationships. User interactions will be facilitated through a specialized hypothesis definition language.