Anomaly detection with CNN autoencoders for Cloud-based AI systems

Abstract

Nowadays, the most important problem is anomaly detection. If anomalies are not detected in time it can cause severe damage to the organization's reputation and revenue. To detect anomalies, organizations use different techniques like Convolutional Neural Network autoencoders which can easily detect anomalies. Other techniques are seasonal ESD, and Seasonal hybrid ESD which use time series to detect anomalies. Machine learning techniques like supervised learning, unsupervised learning, and semi-supervised learning are also implemented in cloud-based AI systems to detect anomalies before any damage is done. Anomalies detection techniques face many challenges like big data, imbalance data, and noise which lead them to detect false anomalies.