Multi-objective Automatic Algorithm Configuration for the Classification Problem of Imbalanced Data

Tari, Sara; Szczepanski, Nicolas; Mousin, Lucien; Jacques, Julien; Kessaci, Marie‐Éléonore; Jourdan, Laëtitia

doi:10.1109/cec48606.2020.9185785

Cited by 3 publications

(2 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…). Thus, some works have exploited this parameter variety, as e.g., in [36,37], where optimized versions of MOLS have been proposed for the imbalanced classification problem by means of the automatic algorithm configuration (AAC) paradigm. Note that we have chosen the best identified parameters for MOLS from this literature, i.e., a initial population of solution of size 100, an archive of max size 500, a restart strategy, a random selection of solution and an exploration of all neighbors.…”

Section: Multi-objective Local Searchmentioning

confidence: 99%

A hybrid CP/MOLS approach for multi-objective imbalanced classification

Szczepanski

Audemard

Jourdan

et al. 2021

Proceedings of the Genetic and Evolutionary Computation Conference

View full text Add to dashboard Cite

In the domain of partial classification, recent studies about multiobjective local search (MOLS) have led to new algorithms offering high performance, particularly when the data are imbalanced. In the presence of such data, the class distribution is highly skewed and the user is often interested in the least frequent class. Making further improvements certainly requires exploiting complementary solving techniques (notably, for the rule mining problem). As Constraint Programming (CP) has been shown to be effective on various combinatorial problems, it is one such promising complementary approach. In this paper, we propose a new hybrid combination, based on MOLS and CP that are quite orthogonal. Indeed, CP is a complete approach based on powerful filtering techniques whereas MOLS is an incomplete approach based on Pareto dominance. Experimental results on real imbalanced datasets show that our hybrid approach is statistically more efficient than a simple MOLS algorithm on both training and tests instances, in particular, on partial classification problems containing many attributes. CCS CONCEPTS• Computing methodologies → Search methodologies; Machine learning approaches;

show abstract

Section: Multi-objective Local Searchmentioning

confidence: 99%

A hybrid CP/MOLS approach for multi-objective imbalanced classification

Szczepanski

Audemard

Jourdan

et al. 2021

Proceedings of the Genetic and Evolutionary Computation Conference

View full text Add to dashboard Cite

show abstract

“…Secondly, we are also going to add the possibility for solvers to communicate their search space in order to improve their diversification potential. Finally, the performance of our parallel MOLS algorithm should also be investigated on other problems, such as the rule mining problem [28]. Indeed, because the evaluation steps of this problem are more expensive, we believe that hybrid methods would be more suitable.…”

Section: Conclusion and Perspectivementioning

confidence: 99%