I have a list of historical timestamps of when a specific event occurred on a website. Currently the timestamp represents a 30 minute window that the event happened within. I am looking to train a model that will predict a time window that the given event is most likely to happen within.
The different events I am trying to predict can have various frequencies such as events that occur every day, week, month, quarter, year. To start off I am interested in looking at events that happen on a daily basis since I can have the most immediate feedback.
For example, I would like to predict when a website will update a given report on their page, I know that this typically happens once per business day. I have a process that will go to the website every 30 minutes to check and see if the report has been posted. Therefore my dataset is a list of timestamps that represent the end time of a 30 minute window the report was published in.
I would like to speed up my process in for a time window the event is most likely to occur so I can receive the report as quickly as possible, without speeding up the process 24 hours per day.
What would be the best way to attack this problem? The prediction will need to adjust every day based upon a new time from the day before.
