I have a bunch of data of customer actions that have taken place at particular times. I was trying to figure out what the first steps would be to build a model that could predict in the future if a customer was going to take an action at some particular time of the day. I could run the model every hour or so and see if a customer was likely to take an action during that hour at that particular time. I also want to use outside temperature and dayOfWeek as variables. Not sure if it will even work or not, but something I'm running into is all I have is postive event data.
I know all of the times a customer didn't take an action, but I don't have records that exist for that, I just have the times a customer did. I can create records with different values of the variables ( like all of the hours or days of the week they didn't take an action ) but does that help in the end and/or is it neccessary? And lastly, I was thinking that a Logistical Regression model might be a good starting choice. But ( clearly ) my ML/AI background is fairly sparse.
I've extracted the data but since it's all positive data I'm not sure it will work; most examples I see use data that has positive and negative outcomes.