How to "count" certain events in a time series


What techniques are suitable and what do I need to learn in order to detect and count the number of "events" (pic) that consist of:

a) shape 1 3 and 4 and not of (2 and 5) or

b) of all shapes above the blue line with a certain minimum "volume" like shapes 1 and 4?

My first idea is to use a window function and somehow integrate the area but i want to be sure to use an appropriate procedure. I havent worked with time series before and I dont know were to get startet. I have to use python for this project.

enter image description here


Posted 2017-04-27T17:37:47.823

Reputation: 23



Try generating a dictionary of patterns you want to identify. You can then use convolutions/ cross-correlations to identify where these patterns appear in your data.

This method is also called 'matched filter'.


Posted 2017-04-27T17:37:47.823

Reputation: 1 329


I suggest to start with "outlier detection", "anomaly detection", filtering methods. Its pretty wide topic to cover but you need to start from somewhere.

Vitaly Portnoy

Posted 2017-04-27T17:37:47.823

Reputation: 61