With R2015a use the xlsread function to import the data. Get all of the outputs in case there is column header information that you need to know in order to use the data correctly. That will pe present in the second and third outputs. Ideally, you should also have a time vector or at least a sampling frequency as well as the input/output data.
Then, depending on the information you have, use the iddata function (or idfrd, whatever is appropriate to your data) and go from there. I use the ssest function to define the system, since it is the most robust (at lelast in my experience).
I know nothing about your data, so explore the documentation to determine how best to characterise and identify your system.