I need some help with agglomerative clustering please
Hello everyone! I am working with RapidMiner for a week now and I cannot figure out how to solve my problem or to be more specific: I need some inspiration for the work with RapidMiner.
Here is my starting point:
have a csv-file which contains several examples of data from sensors of
a fictional production machine. The first row will be a timestamp which contains the time when the sensor collected data. The second one will be the name of the event which happened. Attached you will find some data example as I cannot upload it here.
you can see, from time to time an error has accurred (yellow mark)
which I want to analyse why it happened. The assumption is that events
which happened in a short time before
"error occurred" have a higher possibility to cause this problem. Events
which happened a long time before the error occurred have a lesser
- After doing the tutorial and reading some
questions from the community I decided to try an agglomerative cluster
to cluster all the events which occurred in the time before the event
- That is why I want to take the event
"error occurred" as my zero and measure the time distances between zero
and the events happened before in order to determine which failure of a
sensor will probably lend into the the event "error occurred".
- My thought was to maybe split the data at a first step after each "error occurred" into smaller sub-files and try to apply the agglomerative cluster.
Could you guys please give me an inspiration to solve my problem or could you please tell me if this is possible like I presented my ideas?
Thanks in advance and have a nice week!