Proposing a Neural-Genetic System for Optimized Feature Selection Applied in Medical Datasets

This paper, presents a new system for selecting the best optimized features among a collection of features by combination of neural network and genetic algorithm. Feature selection is an important issue because it has a direct impact on the performance (Specificity, sensitivity) and system efficiency. The proposed system uses neural network for selecting the best features based on Signal to Noise Ratio (SNR), and genetic algorithm for training the neural network by determining the optimum values of weighs and other parameters. This system is a combination of a Multi-Layer Perceptron (MLP) with 3 layers and decimal genetic algorithm We evaluated our proposed system on 10 medical data sets and compared it with binary genetic algorithm that is used widely for feature selection. The results confirmed the superiority of the proposed system in Specificity, sensitivity and the number of selected optimized features.