Context-aware Feature Selection using RNN DAE-RNN DM for Fault Detection in Cloud

Machine Learning (ML) is expected to have an important role in automating the detection of faults in next-generation cloud networks. In ML-based fault detection methods, an efficient method is needed to select the most important features in the presence of missing and/or unknown feature values. In this paper, we propose a context-aware feature selection method, which uses a sensitivity analysis to measure the impact of features on prediction accuracy. Given a time-series, multivariate, and incomplete data, our feature selection method uses Recurrent Neural Networks (RNN) and Auto-Encoder (AE) with an RNN and discriminative model (DM) to estimate the missing features and their sensitivities. The RNN and DAE handle time-series data and missing feature values, respectively, while the RNN DM predicts the system status. Our evaluations on the real-world Google cluster dataset shows that the proposed method outperforms the existing methods in terms of F<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf>-score.