Say I have a list of timestamps which represent completion time for a piece of work. These timestamps are also categorized, but that doesn't necessarily need to be a factor. The piece of data we're trying to predict is the start time. We can generalize that tasks don't overlap. A naive approach would be to take the end time of one task and set that as the start time for the next consecutive task, but we can't assume the end time of one task corresponds to the start time of the next consecutive task.
The user can manually enter actual values when they are wrong, and this can be used for training the data. My data set of trained data would be in the thousands.
I'm wondering if there's any class of algorithms, papers, etc that I could be searching for that'd cover this topic.
[link][2 comments]