I have a data that look like:

| Event | Start time | Duration |
| A     | 0          | 2.5      |
| B     | 0.5        | 2.6      |
| C     | 1.7        | 2        |
| D     | 3          | 1.6      |
| A     | 3.2        | 2.2      |

I need to group these events based on Start time and duration. I have seen time series clustering problems that consider the start time only. How do I do this one?

Edit: It is expected that the events that start and end together is supposed to be in one group. That is if 'A' start at time=0.0 and 'B' at time=0.5. Both have a similar duration also. If this pattern happens every time, we can say that 'A' and 'B' falls into a group.

I do not want any prediction. I just need to group these events.


"You have the data that look like this", Does it mean you have more data than 5 observations? If yes, you can perform KNN, Naive Bayes, LDA. Actually the problem you refers to is quite simple only two predictors.

You set the response as the event type, A B C D, and the predictors as duration and start Time.

Now it is your purpose of solving the problem, are you... 1. create a predictive model the predictive model only wants the correct answer out of the blue so in this in this case, it does not provide the explanatory power for the result. KNN is suited for this purpose

  1. create an inference model so you try to explain the event the impact associated with the variable you use as the predictor and each of the predictor coefficients will give the numeric explanatory value. LDA, Decision tree, Naive Bayes are more appropriate in this case.
  • $\begingroup$ Yes. The data has so many data points. And it is not a prediction problem. Just a clustering only. See the edit. $\endgroup$ – ST Alfas Jan 2 '19 at 8:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.