# Large time series

I have time series data, of some numerical measure. The time interval is in seconds (although I have gaps in the data - missing values). I have data from 2 consecutive months, while for each month I have all the days, and 24 hours a day. For each hour, I have 60 minutes, and for each minute I have values for the seconds. This sums up to a series with over 1 million observations. The question I need to answer is: Given a history of 20 seconds, is it possible to predict the rate of the numerical variable in the next 10 seconds? My variables are: Time variable (date+time), Month, Day, Hour and Minute (all extracted from the time variable), and the numerical variable itself. I am not sure how to approach this problem. Can you give me some advice please? So far I even failed in visually looking at it, the statistical software can't plot it due to the size of it.

• Have you tried inspecting a part of the data? For example the last 5 days? – Euphe Aug 4 '16 at 7:11
• If you are still interested, could you augment this Q with some plots of part of the data, maybe an autocorrelation function for upto 20 seconds? If only interest is in prediction into the next 10 seconds, then maybe the large-scale structure is not so interesting? A link to the data, if possible, would be interesting – kjetil b halvorsen Dec 16 '18 at 2:44