We propose a novel feature selection methodology based on game theory. In this context, the players are the various feature selection methods and the characteristic function (payoff) represents the feature ranking agreement within a coalition of players. The Shapley value assigned to each feature selection method is computed and ranked from higher to lower. The best feature selection method is identified as the one having the highest Shapley value. Finally, we have performed a score fusion scheme using the Borda Count (BC) consensus function as a benchmark to the maximum-Shapley value proposed approach. In order to validate the results obtained experimentally, we have performed a classification using a set of UCI and Statlog datasets by invoking an SVM classifier. Experimental results demonstrate the efficiency of the proposed methodology compared to some state-of-the-art approaches.
We propose a new filter methodology for feature selection using the concept of game theory whereby features are assimilated to players. In this game theoretical context, a strategy corresponds to a particular affinity between a group of features forming a cluster, and the payoff function is computed based on the weighted distance between a feature and a cluster. A zerosum two-player game problem is solved through a global combination of pairwise features. Finally, each feature is represented by the value of the objective function, at the optimal solution, which indicates the contribution of each feature. The importance of features is then evaluated by their optimal values. To validate the effectiveness of the proposed methodology, we have conducted a classification task utilizing SVM on various UCI and statlog datasets. The experimental results show that the proposed scheme leads to improvement in classification performance, when compared to mRMR and Fisher score algorithms.
Efficient classification depends strongly on the quality of the dataset used in experiments. In this paper, we generated a dataset consists of six spectral features extracted from the MSG-SEVIRI satellite images. Each feature represents the brightness temperature of the corresponding pixel. We are based on meteorological radar images acquired in Setif region (Algeria) to assign a class to each feature vector, where we take account of the spatial and spectral resolution difference between radar and satellite images. We are interested to the identification of raining clouds, non-raining clouds and absence of clouds. The application of K Nearest Neighbors (KNN) classifier to the dataset generated performs very well. Using Euclidean metric for classifications, the overall accuracy is 99.46% and the Kappa coefficient attaint 99.13%. In order to validate the results obtained experimentally we have performed an in situ validation using eight ground measurements over the north Algeria. By computing different evaluation measure parameters, experimental resultsdemonstrate the efficiency of the proposed methodology in discriminating between raining and non-raining clouds.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.