# Multivariant time series forcasting

I have a multivariate time series data with Timestamps as( 'season' with four categories: 1,2,3,4, 'month' with 12 categories and 'day' with 5 categories excluding Saturday and Sunday) and I have to predict the target variable 'absenteeism as no of hours'. I have two questions : 1) How shall I deal with the timestamp 2) what category does this problem fall into like classification, regression, clustering? and how can I approach this problem? Any hint would be useful.

• It's unclear what you might mean by "multivariant:" could you clarify? This could be a multivariate problem if you were tracking more than one target variable simultaneously, but it's not. And what distinction are you making between "timestamp" and just time itself? And since you state your interest is prediction, why are you asking about the category of the problem? It's a prediction problem! – whuber Sep 5 '18 at 12:28
• @whuber I meant, I want to convert the above timestamps something like season/Month/Day. So, the first question was meant to understand, what might be the best way to do this? The second question: The target features in the data I have 20 classes, but an employee may be absent for any number of days. So, I cannot take it as a classification problem.So, How shall I approach this problem ? – Akash Dubey Sep 5 '18 at 12:38
• Could you explain what you mean by "20 classes"? You have stated it is a number of hours: that's a single numerical quantity. I also don't follow what you mean by "convert," since you have stated the timestamps already are season, month, and day. It would help to be more specific and less vague in your descriptions. – whuber Sep 5 '18 at 12:43
• @whuber 20 classes mean, the target class has only 20 distinct values as the 'number of hours absent', but still I cannot approach it as a multiclass classification problem or can I ?. Also, Season, Day, Month are in different columns, that's why I asked, how can I deal with them? – Akash Dubey Sep 5 '18 at 12:47