cross flow filtration may be suggested to locate the product (protein ) stream by placing suitable NMWCO. 

There are severals for protein purification, such as ：Ion exchange chromatography，Affinity chromatography，hydrophobic interaction chromatography, exclusion chromatography and electrophoresis.

Try doing a subcellular protein fractionation (you can buy a kit for this). Basically, it enables you to perform a stepwise separation and preparation of cytoplasmic, membrane, nuclear soluble, chromatin-bound and cytoskeletal protein extracts from tissue samples. Extracts obtained are compatible with a variety of downstream applications, including Western blotting and reporter-gene and enzyme-activity assays. 

As previously mentioned, there are some purification methods to localize a protein. Nevertheless, there are many bioinformatics tools which are of use in subcellular localization of proteins based on protein sequence. You may also want to try out these websites provided below:

http://psort.hgc.jp/

http://cello.life.nctu.edu.tw/

http://www.cbs.dtu.dk/services/SignalP/

http://www.cbs.dtu.dk/services/TargetP/

In addition to subcellular fractionation, if you have antibody to the protein, another method is to use the antibody coupled to a microscopy procedure for immunolocalization. Yet another alternative is, if cDNA is available, to express a fusion protein that has a fluorescent tag. If you are working with a well-studied organism, there might even be published information using the fluorescent tag already. http://www.nature.com/nmeth/journal/v10/n4/full/nmeth.2377.html 

Fractionation, immunofluorescence, and fusion proteins are the way to go. Be aware that the localization may be cell type dependent, and that you will often find a distribution throughout the cell. For example, there are proteins that you'll find in the cytoplasm that shuttle into and out of the nucleus, but  you won't find them in the nucleus without inhibiting transport.

If your aim is to purify a specific protein I would recommend to start with a bioinformatic approach as suggested above by Oveis Jamialahmadi. You can also use the SMART algorithm (http://smart.embl-heidelberg.de/) to check for specific domains that can give you a hint regarding the localization (signal peptide, trans membrane domain, DNA binding domain, etc.). Then do sub cellular fractionation (hopefully the bioinformatic analysis could guide you to the right fraction and save you the fractionation kit).

IMHO the first step should be to perform extraction with subsequent activity measurement. If you will be succesful you can further optimize. Roughly, if you can extract it without the use of detergent, it's cytoplasmic. If low amount of detergent are required, it's in some compartment and if high amount of detergent is needed, it is in membrane.

However, if the purification is the sole target, you do not need to bother with this much and keep this for later.

The tangential filtration won't help you, as proteins both from cytosol and membrane can have particular size.