Classification when 80% of my training set is of one class.

I'm new to ML but I've been doing the ML course on coursera and have started messing about with Weka. I'm doing classification on 50,000 records where 79% of my predictor variable, y = 0. Any classification simulations I run also have a 79% success rate. Is my data set too noisy? Is there anything I can do improve this or not?