The subject techniques are not identifying the actual outliers in a frequency dependent data series. The data can be compared to a time series sound data. Sometimes, due to certain artefacts the measured data stored some irrelevant measurements containing spikes. These incorrect measurements are not being detected by the subject techniques. Therefore, I would appreciate suggestions regarding certain alternate highly effective automated outlier detection techniques. I am open for suggestions for both Python and Matlab implementation. Thanks!
If your data points are more you can use density based clustering algorithm (DBSCAN, HDBSCAN) it will separate the outliers.
Thanks Gare Suman, I will try this technique and update if the mentioned approach is able to detect the outliers.
Meanwhile I am open for more suggestions from readers.
I noticed that filloutliers command in Matlab does a reasonably good job with my measured data. For future reference of readers I am placing the relevant command as follows:
[B, TF]= filloutliers(Data,'center','median','ThresholdFactor',1.5)
Keeping in view the normal distribution of the considered data, the command detects the index number (TF) and values of the outliers (B).
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Cell Biology
- Culture Cells
- Cell Line