Pretty vague. Starting from
here, a discrete form might be:
[$]r_t = r_{t-1} + \theta_t + \epsilon_t[$], where [$]\epsilon_t \sim N(0,\sigma^2)[$], where
'[$]\sim[$]' means "is distributed as".
N(0,v) denotes an (independent) normal distribution at each step with mean-0 and variance [$]v[$]. Finally, [$]\theta_t[$] is simply some deterministic sequence.
As far as your application, hard to say, other than to learn how to apply maximum likelihood to your data. There are a zillion books on time series modelling.