I'm not an expert in this nor a lawyer. My understanding is you can't patent an abstract mathematical algorithm in its own right. What you might be able to patent is the application of an algorithm to solving a specific problem. This is a bit of a moving target as there have been some recent legal decisions with regards to software patents.
With regards to whether you should pursue a patent, understand that getting a patent takes time and money. If you are in academia, it might be more useful to you personally to publish your algorithm. Also if you are associated with an university, the university may be able to provide guidance with respect to patenting and licensing.