An evolutionary algorithm for LTS-Regression: A comparative study

Mutation: replace one of the p points by one of the n− p other points. Move: parallel shift of the estimated hyperplane, such that it hits a random one of the n-p remaining points. π is controlled with the parameter percentagemove PER. Simulated Annealing ANN for not getting trapped in a local optimum. Intercept adjustment ADJ possible for reducing the LTS–value further. Further parameters: waitforimprovement WFI, which is controlling nW, total number of generations GEN and total number of iterations ITE. 4.) Designed Experiments for comparison of algorithms

[1]  T. Bernholt.  Robust Estimators are Hard to Compute , 2006 .

[2]  Peter J. Rousseeuw,et al.  Robust Regression and Outlier Detection , 2005, Wiley Series in Probability and Statistics.

[3]  PETER J. ROUSSEEUW,et al.  Computing LTS Regression for Large Data Sets , 2005, Data Mining and Knowledge Discovery.