Anomaly Detection for Multiple Variables + Score

by Sandy   Last Updated January 14, 2018 08:19 AM

I am working on a project which involves detection of a motor failure. My dataset involves data of different attributes of motor, such as speed, vibration, acceleration captured at seconds level. The motor runs every day for a short period of time (ranges from 6 min to 30 min). I have data from 30 days. I need to find events of motor failure in these 30 days and score every event.

Sample Data

Day      Time                   Speed      Vibration      Acceleration
 1       2015-09-22 14:15:26     40           3.2             0.42
 1       2015-09-22 14:15:27     41           3.3             0.44
 1       2015-09-22 14:15:28     42           3.4             0.45
 1       2015-10-21 20:37:48     49           2.2             0.46
 1       2015-10-21 20:37:49     49           2.2             0.47

I am able to find outliers/anomalous readings for each column using graphs;but I am not sure how to detect anomaly considering all/some variables. Could anyone help me in solving this problem and guide me in the correct direction?



Related Questions



Training of Isolation Forest with "normal" data only?

Updated September 24, 2018 16:19 PM

Is a one class naive bayes possible?

Updated February 28, 2017 23:19 PM