# How to find similarity of two series over time containing periodic trends?

Considering the data is received from a streaming source each second.How to distinguish if both the line graphs 'look' same/different in real time, statically, like the picture given below Edit:
1.Not sure if there is something as real-time periodic correlation mechanism then cross-correlation would perhaps be an ideal solution ?
2. Comparing slope of the two Line is the last option I would go with. 3. If statistically there is no way to solve this then I would look at machine learning to solve this

Maybe it is a huge oversimplification, but you can try: What if you do just arithmetic difference between the two signals (or divide one signal to another one) and look at the result?

I would expect that you will have spikes where signals become very different and you can make a threshold to find where they are on time.

I am unsure on how to do this statistically apart from using the slope. However, there are some interesting ways to think about time series in a given time period to find similarity. You could do as follows:

1. Use the characteristics of the signal in a given time window and construct an n dimensional vector.
2. You could use similarity measures or distance measures like cosine, manhattan etc to evaluate the similarity between them.
• even with slope methods ,considering both the signals donot rise or fall by equal slopes , how to detect if they are similar? Nov 15, 2018 at 18:47
• I am not quite sure, maybe finding the average of the slope and comparing helps. Here is how you could find the average of the slope : searchink.atlassian.net/wiki/spaces/PROJ/pages/282329252/… I still think finding a vector representation and using other metrics/measures will help you evaluate this better. Nov 15, 2018 at 18:58

be series1 = S and series2 = T

It would come from:

Tn - Sn-1 = Sn - Tn-1, then:

Tn - Tn-1
-------------- = 1
Sn - Sn-1


In this way, favorable cases will give 1 and unfavorable cases will show noise.

• can you elaborate this? Apr 16, 2019 at 8:56
• Of course, the only way to know that the graphs behave the same is by comparing with the previous data. Apr 16, 2019 at 20:57