Bund, Tom and Gataveckas, Tomas and Grunert, Hannes (2026) I choose you... Improving Genetic Algorithms by Association Rules. In: European Conference on Data Analysis 2026, 09-11 Sep 2026, Stralsund, Deutschland. (In Press)
Full text not available from this repository.Abstract
Genetic Algorithms (GA) offer potential for discovering new solutions to complex optimization problems, but their effectiveness depends on the size and structure of the search space. To reduce the search space and guide the evolutionary process, we integrate association rules mined from transactions into the GA. In this way, the algorithm is nudged by favoring meaningful co-occurrences. We evaluate the approach in the context of InSleeve, an online store for Trading Card Games (TCGs). Our approach follows a four-step pipeline: (1) We retrieve cards from the public Pokémon API. In parallel, we collect reference decks from Limitless TCG. (2) Using these test decks, we derive statistics to define constraints for deck composition and to identify frequent combinations. (3) Association rules are mined, focusing on support, confidence and lift to balance rule quality and memory requirements. (4) The GA then iteratively applies recombination, semi-random substitutions, fitness estimation and selection. The pipeline produces a filtered card pool and a curated dataset of reference decklists. From the test decks, we estimate composition ranges of different card types as constraints in the evolutionary search to reduce implausible candidates early and improve the comparability of generated decks. The association-rule component yields co-occurrence patterns that are injected into the GA by influencing initialization and mutation choices to reduce random exploration and to increase synergies.
| Item Type: | Conference or Workshop Item (Poster) |
|---|---|
| Subjects: | Autorenart > Studentische Arbeiten > Seminararbeit Forschungsthemen > Big Data Analytics Autorenart > DBIS-Publikationen Autorenart > Studentische Arbeiten |
| Depositing User: | Dbis Admin |
| Date Deposited: | 16 Jun 2026 10:40 |
| Last Modified: | 16 Jun 2026 10:43 |
| URI: | https://eprints.dbis.informatik.uni-rostock.de/id/eprint/1155 |
Actions (login required)
 |
View Item |
