Análisis Estadístico de Algoritmos Evolutivos para el problema de Selección de Variables

• Jimenez, Victor Adrian ^[1] ; Lizondo, Diego Fernando ^[1] ; Will, Adrián
  1. [1] Universidad Tecnológica Nacional

     Universidad Tecnológica Nacional

     Argentina

     
• Localización: Ciencia y tecnología, ISSN 1850-0870, ISSN-e 2344-9217, Nº. 19, 2019, págs. 65-84
• Idioma: español
• DOI: 10.18682/cyt.v19i19.934
• Enlaces
  • Texto completo (pdf)
• Resumen
  • español

    Durante décadas de investigación en problemas de optimización, se han desarrollado innumerables algoritmos, tanto determinísticos como heurísticos. Aun así, debido a este amplio abanico de posibilidades resulta una tarea compleja determinar cuál de ellos es más adecuado para un problema específico. Se propone en este trabajo una comparativa entre diferentes algoritmos heurísticos de optimización haciendo uso de test estadísticos. Se consideró Simulated Annealing (SA), Algoritmo Genético Simple (sGA), Algoritmos Genético Compacto (cGA) y Deterministic Crowding (DC) aplicados al problema de Selección de Variables para Estimación utilizando Regresión Lineal. Se utilizaron tres casos de prueba, que consisten en la estimación de Radiación Solar, Consumo Eléctrico en la provincia de Tucumán (Argentina), y el tiempo de reaparición de células cancerígenas. Se comprobó que existe suficiente evidencia estadística para afirmar que los algoritmos arrojan resultados significativamente diferentes. Además se concluye que sGA y DC son los algoritmos más apropiados para estos problemas, permitiendo obtener valores de fitness similares siendo sGA ligeramente superior.

  • English

    Decades of research in optimization problems have generated a considerable number of algorithms, both deterministic and heuristic. However, due to this wide range of possibilities, determine which one is the most appropriate for a specific problem is a complex task. In this paper, a comparison among different heuristic optimization algorithms using statistical test is proposed. Simulated Annealing (SA), Simple Genetic Algorithm (sGA), Genetic Algorithms Compact (cGA) and Deterministic Crowding (DC) were used, applied to the Variable Selection for estimation problem using Linear Regression. Three test cases were used: solar radiation at the province of Tucuman (Argentina), power consumption estimation in the same area, and estimation of the reappearance of cancer cells. We concluded that there is sufficient statistical evidence to affirm that the algorithms yield significantly different results. Also, we concluded that sGA and DC were the most suitable algorithms, obtaining similar fitness values, being sGA slightly better.

• Referencias bibliográficas
  • Ahn, C. W., & Ramakrishna, R. S. (2003). Elitism-based compact genetic algorithms. Evolutionary Computation, IEEE Transactions on, 7(4),...
  • Coffin, M., & Saltzman, M. J. (2000). Statistical analysis of computational tests of algorithms and heuristics. INFORMS Journal on Computing,...
  • De Jong, K. A. (1975). Analysis of the behavior of a class of genetic adaptive systems (PhD Thesis). University of Michigan.
  • Denardo, E. V. (2012). Dynamic programming: models and applications. Courier Corporation.
  • Derrac, J., García, S., Molina, D., & Herrera, F. (2011). A practical tutorial on the use of nonparametric statistical tests as a methodology...
  • Eldén, L. (2007). Matrix Methods in Data Mining and Pattern Recognition. Philadelphia, PA, USA: Society for Industrial and Applied Mathematics.
  • Harik, G. R., Lobo, F. G., & Goldberg, D. E. (1999). The compact genetic algorithm. Evolutionary Computation, IEEE Transactions on, 3(4),...
  • Hauschild, M., & Pelikan, M. (2011). An introduction and survey of estimation of distribution algorithms. Swarm and evolutionary computation,...
  • Jimenez, V. A., Lizondo, D., Will, A., & Rodriguez, S. (2017). Short-Term Load Forecasting for Low Voltage Distribution Lines in Tucumán,...
  • Kirkpatrick, S. (1984). Optimization by simulated annealing: Quantitative studies. Journal of statistical physics, 34(5-6), 975–986. https://doi.org/10.1007/BF01009452
  • Kotu, V., & Deshpande, B. (2015). Chapter 12 - Feature Selection. En V. Kotu & B. Deshpande (Eds.), Predictive Analytics and Data...
  • Mahfoud, S. W. (1995). Niching methods for genetic algorithms. Urbana, 51(95001), 62–94.
  • Parsopoulos, K. E. (2010). Particle swarm optimization and intelligence: advances and applications: advances and applications. IGI global.
  • Peng, X., Bessho, M., Koshizuka, N., & Sakamura, K. (2014). A framework for peak electricity demand control utilizing constraint programing...
  • Talbi, E.-G. (2009). Metaheuristics: from design to implementation (Vol. 74). John Wiley & Sons.
  • Varmuza, K., & Filzmoser, P. (2008). Introduction to multivariate statistical analysis in chemometrics. CRC press.
  • Will, A., Bustos, J., Bocco, M., Gotay, J., & Lamelas, C. (2013). On the use of niching genetic algorithms for variable selection in solar...
  • William H. Wolberg, O. L. M. (1992). Wisconsin Breast Cancer Database [https://archive.ics.uci.edu/ml/datasets/Breast+Cancer+Wisconsin+...
  • Williams, M., & Daneshyari, M. D. (2016). Clonal vs. Negative Selection in Artificial Immune Systems (AIS).
  • Yu, X., & Gen, M. (2010). Introduction to Evolutionary Algorithms. En Decision Engineering. Springer.