A feature selection method based on information theory and genetic algorithm

Dealing with the high dimensional datasets, dimension reduction as a pre-processing approach can assist to provide high accuracy, efficiency and scaling procedure particularly in classification problems. In this study, an algorithm for feature selection based on the information theory has been proposed focusing on the dimensionality reduction in classification task. In this approach mutual information between candidate features and label class is measured by considering a new optimal metric. Next to the new MI metric, the meta heuristic algorithm based on genetic algorithm has been applied to increase the speed and efficiency of the proposed method. This approach is applied on the datasets with different dimensions from 13 to 60. The evaluation results show the promising results in term of classification accuracy in comparison with other similar methods. the proposed method has been studied with the mRMR, DISR, JMI, NJMIM data based and the gap between this data contrasted with proposed algorithm.

Dealing with the high dimensional datasets, dimension reduction as a pre-processing approach can assist to provide high accuracy, efficiency and scaling procedure particularly in classification problems. In this study, an algorithm for feature selection based on the information theory has been proposed focusing on the dimensionality reduction in classification task. In this approach mutual information between candidate features and label class is measured by considering a new optimal metric. Next to the new MI metric, the meta heuristic algorithm based on genetic algorithm has been applied to increase the speed and efficiency of the proposed method. This approach is applied on the datasets with different dimensions from 13 to 60. The evaluation results show the promising results in term of classification accuracy in comparison with other similar methods. the proposed method has been studied with the mRMR, DISR, JMI, NJMIM data based and the gap between this data contrasted with proposed algorithm.

Dealing with the high dimensional datasets, dimension reduction as a pre-processing approach can assist to provide high accuracy, efficiency and scaling procedure particularly in classification problems. In this study, an algorithm for feature selection based on the information theory has been proposed focusing on the dimensionality reduction in classification task. In this approach mutual information between candidate features and label class is measured by considering a new optimal metric. Next to the new MI metric, the meta heuristic algorithm based on genetic algorithm has been applied to increase the speed and efficiency of the proposed method. This approach is applied on the datasets with different dimensions from 13 to 60. The evaluation results show the promising results in term of classification accuracy in comparison with other similar methods. the proposed method has been studied with the mRMR, DISR, JMI, NJMIM data based and the gap between this data contrasted with proposed algorithm.

Dealing with the high dimensional datasets, dimension reduction as a pre-processing approach can assist to provide high accuracy, efficiency and scaling procedure particularly in classification problems. In this study, an algorithm for feature selection based on the information theory has been proposed focusing on the dimensionality reduction in classification task. In this approach mutual information between candidate features and label class is measured by considering a new optimal metric. Next to the new MI metric, the meta heuristic algorithm based on genetic algorithm has been applied to increase the speed and efficiency of the proposed method. This approach is applied on the datasets with different dimensions from 13 to 60. The evaluation results show the promising results in term of classification accuracy in comparison with other similar methods.