Reference Paper : Unsupervised Anomaly Detection via Variational Auto-Encoder for Seasonal KPIs in Web Applications – https://arxiv.org/abs/1802.03903
Fetus Anomaly detection using Variational Auto Encoder
https://physionet.org/content/adfecgdb/1.0.0/ -Dataset used for modeling
Since the data format is in edf, it needs to be converted to csv, which is done with the mne library
fetalAnomalyAbdomen.py - Determines fetal heart rate anomaly from electrocardiogram signals captured from maternal abdomen fetalAnomalyBrain.py - Identifies anomaly from electrocardiogram signals captured from fetal brain
Pre-requisites :
pip install donut
pip install tfsnippet
pip install mne
Since data was initially not labelled, the initial labelling of anomalous electrocardiogram signals was done by considering signals >= .00025 and signals <= -.00025. However the same could be accomplished by using a single class SVM.