Accelerating a multi-objective memetic algorithm for feature selection using hierarchical k-means indexes
| Autor: | Pablo Moscato, Regina Berretta, Francia Jiménez, Claudio Sanhueza |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Multivariate statistics
021103 operations research Computer science 0211 other engineering and technologies k-means clustering Feature selection 02 engineering and technology computer.software_genre Multi-objective optimization Exponential growth Feature (computer vision) 0202 electrical engineering electronic engineering information engineering Memetic algorithm Combinatorial search 020201 artificial intelligence & image processing Data mining computer |
| Zdroj: | GECCO (Companion) |
| DOI: | 10.1145/3205651.3205707 |
| Popis: | The (α, β)-k Feature Set Problem is a mathematical model proposed for multivariate feature selection. Unfortunately, addressing this problem requires a combinatorial search in a space that grows exponentially with the number of features. In this paper, we propose a novel index-based Memetic Algorithm for the Multi-objective (α, β)-k Feature Set Problem. The method is able to speed-up the search during the exploration of the neighborhood on the local search procedure. We evaluate our algorithm using six well-known microarray datasets. Our results show that exploiting the natural feature hierarchies of the data can have, in practice, a significant positive impact on both the solutions' quality and the algorithm's execution time. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|