CODEWITHSUNDEEP
machine-learningcomputer-visioneuclidean-distancecosine-similarity
user2359877
user2359877

Reputation: 45

Similarity Metrics

I am trying to research on different metrics and found many ssimilarity metrics : Euclidean distance Dynamic Time Warping, Edit Distance with Real Penalty DISSIM , Sequence Weighted Alignment model, Spatial Assembling Distance.

However I had a question : Is there any other metric that gives similarity in a time series manner. I am looking for something that not only calculates similarity metric between the features but also arranges them in a Time series manner (in the order the features are received)?

For eg: if we have features F1,F2,F3,F4,F5,F6,7,F8 and of them if features F1 F2 F4 F6 F8 are similar

Is there an algorithm that gives an output F1 F2 F4 F6 F8 are similar in this order :. ie is there an algorithm that gives result in a series manner (in terms of time )

Thankyou!

Upvotes: 0

Views: 379

Answers (1)

eamonn
eamonn

Reputation: 11

The question is confusing. But for what it is worth...

Dynamic Time Warping is NOT a metric. Spatial Assembling Distance is not a metric.

You should read http://www.cs.unm.edu/~mueen/DTW.pdf

If you are doing time series similarity comparisons, then i am 99% sure your should use DTW. Source, I have tested more distance measures on more datasets, than the rest of the world combined. http://www.cs.ucr.edu/~eamonn/time_series_data/

Upvotes: 1

Related Questions