Optimizing a Deep Learning Model in Order to Have a Robust Neural Network Topology

Abstract

Abstract-In this study, a method based on different feature engineering / feature extraction / feature derivation is proposed for improving air passenger forecasting by machine learning existing libraries. In this kind of formulation, we kept focus on creating different kinds of datasets that differ one from another by methodology so we extracted new features and compared new feature space with original feature space in terms of variable importance. We conducted experiments to improve the variance by aggregating all the features in final feature space. Finally, we optimized a deep learning model to have a Robust Neural Network Topology.Index Terms-Deep learning, logistic regression, feature extraction, neural network.