I am stuck on what should i use to detect anomalies in time series data.
The problem is:
Suppose I have the timeseries data of products of an e commerce website. I also have a list of similar products. I want to find out when i get an anomaly in the price of any product.
Example scenario: If the website puts price of LG TV as irregularly low or high, but the competitor phones prices change are not so drastic, the model should predict that an anomaly has occurred.
Should i use unsupervised clustering algorithm to detect this?
Or a supervised RNN (But the anomaly data is very rare and this will be hard to train).
Or is there any other method you would suggest?