The forthcoming Dark Energy Spectroscopic Instrument (DESI) experiment plans to measure the effects of dark energy on the expansion of the Universe and create a 3D map of the Universe using galaxies up to z~1.6 and QSOs up to z~3.5.In order to create this map, DESI will obtain spectroscopic redshifts of over 30 million objects; among them, a majority are [OII] emitting star-forming galaxies known as emission-line galaxies (ELGs). These ELG targets will be pre-selected by drawing a selection region on the g-r versus r-z colour-colour plot, where high-redshift ELGs form a separate locus from the lower redshift ELGs and interlopers. In this paper, we study the efficiency of three ELG target selection algorithms - the Final Design Report (FDR) cut based on the DEEP2 photometry, Number Density Modelling (NDM) and Random Forest - to determine how the combination of these three algorithms can be best used to yield a simple selection boundary that will be best suited to meet DESI's science goals. To do this, we selected 17 small patches in the DESI footprint where we run the three target selection algorithms to pre-select ELGs based on their photometry. We observed the pre-selected ELGs using the MMT Binospec, which is similar in functionality to the DESI instrument, to obtain their spectroscopic redshifts and fluxes of 1054 ELGs. By analysing the redshift and fluxing distribution of these galaxies, we find that although NDM performed the best, simple changes in the FDR definition would also yield sufficient performance.