Meta-Learning-Based Algorithm Selection for Multi-Objective Ridesharing-Inspired Optimization

• Tarcisio A. P. Brito UFBA
• Juvenal B. A. Silva UFBA
• Islame F. C. Fernandes UFBA

DOI: https://doi.org/10.5753/eniac.2026.29840

Resumo

Algorithm selection aims to identify the most suitable optimizer for a given instance, but remains challenging in multi-objective combinatorial optimization because performance depends on Pareto set characteristics. This paper investigates meta-learning for algorithm selection in a three-objective ridesharing-inspired problem integrating route planning, passenger assignment, and reward collection. We use this problem as a structured testbed derived from the Traveling Salesman Problem. We compare static, Pareto-based landscape, and combined meta-feature representations using 252 synthetic instances and three multi-objective evolutionary algorithms: MOEA/D, SPEA2, and IBEA. A multi-output Random Forest model is trained to predict algorithm performance under HV and IGD indicators. Results show that static features are highly informative, while Pareto-based landscape features alone provide limited predictive power. Combined features consistently outperform the single-algorithm baseline, but the marginal gain over static-only meta-features is limited and indicator-dependent.

Referências

Brazdil, P., Van Rijn, J. N., Soares, C., and Vanschoren, J. (2022). Metalearning: Applications to automated machine learning and data mining. Springer Nature.

Cenikj, G., Nikolikj, A., Petelin, G., Van Stein, N., Doerr, C., and Eftimov, T. (2026). A survey of features used for representing black-box single-objective continuous optimization. Swarm and Evolutionary Computation, 101:102288.

Cosson, R., Derbel, B., Liefooghe, A., Aguirre, H., Tanaka, K., and Zhang, Q. (2021). Decomposition-based multi-objective landscape features and automated algorithm selection. In European Conference on Evolutionary Computation in Combinatorial Optimization (Part of EvoStar), pages 34–50. Springer.

Cosson, R., Derbel, B., Liefooghe, A., Verel, S., Aguirre, H., Zhang, Q., and Tanaka, K. (2022). Cost-vs-accuracy of sampling in multi-objective combinatorial exploratory landscape analysis. In Proceedings of the Genetic and Evolutionary Computation Conference, pages 493–501.

Daolio, F., Liefooghe, A., Verel, S., Aguirre, H., and Tanaka, K. (2017). Problem features versus algorithm performance on rugged multiobjective combinatorial fitness landscapes. Evolutionary Computation, 25(4):555–585.

Li, W., Özcan, E., John, R., Drake, J. H., Neumann, A., and Wagner, M. (2017). A modified indicator-based evolutionary algorithm (mIBEA). In IEEE Congress on Evolutionary Computation (CEC), pages 1047–1054. IEEE.

Liefooghe, A., Daolio, F., Verel, S., Derbel, B., Aguirre, H., and Tanaka, K. (2019). Landscape-aware performance prediction for evolutionary multiobjective optimization. IEEE Transactions on Evolutionary Computation, 24(6):1063–1077.

Liefooghe, A., Vérel, S., Lacroix, B., Zavoianu, A.-C., and McCall, J. (2021). Landscape features and automated algorithm selection for multi-objective continuous optimisation problems. In GECCO, pages 421–429.

López-Ibáñez, M., Dubois-Lacoste, J., Cáceres, L. P., Birattari, M., and Stützle, T. (2016). The irace package: Iterated racing for automatic algorithm configuration. Operations Research Perspectives, 3:43–58.

Marques, T. S., Cezario, S. F., Goldbarg, E. F. G., Goldbarg, M. C., and Maia, S. M. D. M. (2019). Quota traveling salesman with passengers and collection time. In 2019 8th Brazilian Conference on Intelligent Systems (BRACIS), pages 299–304. IEEE.

Ochoa, G., Liefooghe, A., and Verel, S. (2024). Funnels in multi-objective fitness landscapes. In International Conference on Parallel Problem Solving from Nature (PPSN), pages 343–359. Springer.

Rice, J. R. (1976). The algorithm selection problem. Advances in Computers, 15:65–118.

Silva, B. C., Fernandes, I. F., Goldbarg, M. C., and Goldbarg, E. F. (2020). Quota travelling salesman problem with passengers, incomplete ride and collection time optimization by ant-based algorithms. Computers & Operations Research, 120:104950.

Silva, J. B. A., Brito, T. A. P., Rios, R. A., Rios, T. N., and Fernandes, I. F. C. (2025). Multi-objective traveling salesman problem with profits and passengers: Exact, heuristic and evolutionary approaches. In Proceedings of the Genetic and Evolutionary Computation Conference Companion, pages 419–422.

Verel, S., Liefooghe, A., Jourdan, L., and Dhaenens, C. (2011). Pareto local optima of multiobjective NK-landscapes with correlated objectives. In European Conference on Evolutionary Computation in Combinatorial Optimization, pages 226–237. Springer.

Zhang, Q. and Li, H. (2007). MOEA/D: A multiobjective evolutionary algorithm based on decomposition. IEEE Transactions on Evolutionary Computation, 11(6):712–731.

Zitzler, E., Laumanns, M., and Thiele, L. (2001). SPEA2: Improving the strength pareto evolutionary algorithm. TIK Report, 103.