This is a big research project. For a school science fair project, would just a demonstration of the idea be enough?
Think about breaking the project into sections:
a) obtain accurate measurements of the Earth's magnetic field over a long time for many locations where earthquakes have happened,
b) mark when earthquakes happened, and look for similar changes in the magnetic field measurements before
eathquakes in different places,
c) write a Mathematica code that filters data for that sort of change.
The data for (a) is probably available, but you might have to write a lot of letters asking how to obtain copies.
The author of the paper at http://cty.jhu.edu/imagine/docs/Sensing_Danger.pdf might be willing to give suggestions,
or the references at the end might point to web sites with it.
A lot of people are working on (b) already.
A demonstration of (c) might be enough. Take sample data from a paper and show how you could look for unusual changes.