In their 2008 report, the Surveys state that there is a 99 (99.7 actually)% chance that there will be at least one earthquake of magnitude 6.7 in the California area over the next 30 years, and a 46% chance that there will be one over 7.5 (which will probably be in Southern California).
But how does one come up with an earthquake prediction?
1) Geodetic (relative movement of tectonic plates) -- GPS satellites can track California plate movements. *Here, you can measure tectonic movements with GPS triangulation yourself, thanks to Penn State.
2) Geological -- Rocks along faults can show distance of slippage.
3) Seismological and Paleoseismological
In 2007 scientists making up the Working Group on California Earthquake Probabilities (WGCEP, just because we always need more acronyms) started the new Uniform California Earthquake Rupture Forecast (UCERF). The forecast utilizes years of data to come up with best guesses as to the likelihood/locations of big quakes in both our near and distant futures.
*Because the UCERF was reviewed multiple times by independent scientific review panels as well as by the California and National Earthquake Prediction Evaluation Councils, the Geological Surveys call it "one of the most extensively reviewed earthquake forecasts ever produced."
Individual Faults, and Probability of >6.7 quake (from the '08 report):
Southern San Andreas 59%
Hayward-Rodgers Creek 31%
San Jacinto 31%
Northern San Andreas 21%
* See the "Executive Summary" of the forecast for full details.