Taining LIBSVM with time series data

Taining LIBSVM with time series data

I am working time series problem for several months and still unable to
figure it. Here is the question. I have data collected for 6 days, and the
collection starts from 8:00 am to 11:00. So the data looks something like
this
var1 | var2 | var3
1 2.2 UT1
2 3 UT2
3 1.2 UT3
1 2.2 UT4
2 3 UT5
3 1.2 UT6
1 2.2 UT7
2 3 UT8
3 1.2 UT9
UT represents unixtimestamp. So I converted the var3 into tow columns day
and time. So now the table looks like this
var1 | var2 | var3 | var4
1 2.2 12 8.0
2 3 12 8.25
3 1.2 12 8.5
1 2.2 13 8.0
2 3 13 8.25
3 1.2 13 8.5
1 2.2 14 8.0
2 3 14 8.25
3 1.2 14 8.5
var3 is the day of the month I collected the data and I collected the data
for 6days from 12 to 17. and var4 is time which I collected starting from
8:00 to 11:00 AM for every 20 minutes. The output I am having is a
dependent variable on these variables. So if I need to perform regression
using LIBSVM how can I do that with these?? Do I need to input var3 and
var4 differently? or how does LIBSVM treat this data? DO I need to
normalize the data? I would really appreciate your help.