# HMM - Matlab for data set to detect anomaly

I have a dataset of oil temperatures. The time series consist of 100 hours of measurement at every second. There is an anomaly in the data that I would like to detect using Hidden Markov Models (HMM). In Matlab I used hmmgenerate, however I'm not able to link my data to the assumed probabilities. Instead random states and sequences are generated. Could anyone help me?

• I'm not familiar with Matlab HMM, but from looking at the documentation, perhaps hmmtrain would better suit your objective? mathworks.com/help/stats/hmmtrain.html Feb 3 '17 at 16:02
• Welcome to DataScience SE. If you can post some code and some output it would be helpful for people. Feb 3 '17 at 20:48