Say I have a data set such as the following:
person, Time, Value, Event person1, 2010-07-02 00:00:00, 5.4, 0 person2, 2010-07-02 10:00:00, 12.7, 0
We have a current model in place at work that doesn't take into account the temporal aspect of our data. In that implementation, the model was trained with only unique values for 'person', and it throws away the time variable. However, it has come to our attention that we can look at our data as a sequence instead. This starting time is unique for each person, and clearly associated with only that person, so merely pretending each person is independent and just treating each row as an individual data point wouldnt make any sense. The following is what I've restructured the data as:
person, Time, Value, Event person1, 2010-07-02 00:00:00, 5.4, 0 person1, 2010-07-02 00:00:15, 3.6, 0 person1, 2010-07-02 00:00:30, 2.4, 0 person2, 2010-07-02 10:00:00, 12.7, 0 person2, 2010-07-02 10:01:15, 12.8, 0 person2, 2010-07-02 10:01:30, 13.1, 1
This sequence for each person would continue until and 'event' or 'non-event'. I'm totally unfamiliar with machine learning on time series data. All of the examples I've read with different models treat the data as one big sequence corresponding to one entity, while our data clearly doesn't work like that. Is the way I've structured the data the right way to approach a time series model? And if so, what would be an appropriate model to consider?