Forward Feature Selection Based on Approximate Markov Blanket
Feature selection has many applications in solving the problems of multivariate time series . A novel forward feature selection method is proposed based on approximate Markov blanket. The relevant features are selected according to the mutual information between the features and the output. To identify the redundant features, a heuristic method is proposed to approximate Markov blanket. A redundant feature is identified according to whether there is a Markov blanket for it in the selected feature subset or not.The simulations based on the Friedman data, the Lorenz time series and the Gas Furnace time series show the validity of our proposed feature selection method.
KeywordsFeature Selection Redundancy Analysis Markov Blanket Mutual Information
Unable to display preview. Download preview PDF.
- 6.Koller, D., Sahami, M.: Toward optimal feature selection. In: Proc. Int. Conf. on Machine Learning, pp. 284–292. Morgan Kaufmann, San Francisco (1996)Google Scholar
- 8.Herrera, L.J., Rubio, G., Pomares, H., Paechter, B., Guillén, A., Rojas, I.: Strengthening the Forward Variable Selection Stopping Criterion. In: Alippi, C., Polycarpou, M., Panayiotou, C., Ellinas, G. (eds.) ICANN 2009. LNCS, vol. 5769, pp. 215–224. Springer, Heidelberg (2009)CrossRefGoogle Scholar