Detection of One Dimensional Anomalies Using a Vector-Based Convolutional Autoencoder

摘要: Anomaly detection is important to significant real life entities such as network intrusion and credit card fraud. Existing anomaly methods were partially learned the features, which not appropriate for accurate of anomalies. In this study we proposed vector-based convolutional autoencoder (V-CAE) one dimensional detection. The core our model a linear autoencoder, used construct low-dimensional manifold feature vectors normal data. At same time, neural (V-CNN) extract features from vector data before after that makes deep efficient This unsupervised learning method only in training phase. We combined abnormal score calculated two reconstruction errors: (i) error between input output whole architecture (ii) encoder. Compared with nine state-of-the-arts methods, V-CAE shows effective stable results AUC 0.996 estimating anomalies based on several benchmark datasets.