A Genetic Algorithm for Efficient Descriptive Pattern Mining
Un algorithme génétique pour l'extraction efficace de motifs descriptifs
Résumé
Traditional pattern mining algorithms can find too many patterns in large datasets, including many redundant patterns. Minimum Description Length (MDL)-based methods address this issue through data compression to capture small sets of patterns that capture significant information. However, MDL-based pattern mining techniques, are computationally expensive. As a solution, this paper investigates the efficacy of integrating genetic algorithms (GAs) with MDL-based methods for improved pattern compression in a novel approach called GA4PC (GA for MDL-based pattern compression). By relying on GAs, GA4PC offers a dynamic and adaptable framework for compression tasks. Through experiments, this study reveals that using a GA for MDL-based pattern compression surpasses previous methods in compression ratios, achieves reduced computational time, and produces compressed representations that capture more representative itemsets within the data.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |