-1

I have faced problem with finding the similarity scores between two patterns. For example, I have normal ECG pattern,

enter image description here

and abnormal ECG pattern

[enter image description here][2].

Then I want to get find the accuracy of normal pattern while comparing with abnormal ECG pattern. So my data consists of ID, datetime, Seq and Value. First I was thinking of using time series approach. However, it doesn't solve my problem. So could anyone help me with that?

Joe Ferndz
  • 8,417
  • 2
  • 13
  • 33
M_K
  • 1
  • 1

1 Answers1

0

I'm not completely sure i understood the question but i can't coment right now.

If you want to get the accuracy of the second plot comparing to the first plot, there are many ways, a simple and usefull one is the mean absolute error were you add all the values \delta = \sum (x_expected - x_obtain)^2 but this accuracy methods depends on how you want to compare.

Here in the tensorflow.keras documentation you can find many diferent metrics to decide how to compare the graphs.

Tomas Crosta
  • 1
  • 3
  • thank you. Actually I take normal pattern as training data, then I take abnormal pattern as test data. Then I will compare test pattern according to the training data. If the test data is similar with train data, then the algo should give similarity score. – M_K Mar 19 '21 at 04:36
  • That's the thing, there is actually no perfect answer so you have to choose, for me the easiest to implement is using the mean squear error but you can try different things like fitting big polinomials and compare the coeficients. Good luck – Tomas Crosta Mar 19 '21 at 12:42