I am working to find anomaly/outliers in sensor data using unsupervised machine learning (without training dataset). I have around 20000 samples taken per minute of various sensors. I just need to detect the outliers(point, collective and contextual). I have been through many articles but not getting the right link.
clustering is meaningless and suppose one reading (500) during the day is normal but during night it is an outlier.
I am doing in python, already done with preprocessing of data using Knn Imputation.
can someone tell me what should I do next, all the algorithms I find have some or the the other flaws!