We describe a method for predicting detection limits of minority elements in electron energy loss spectroscopy (EELS), and its implementation as a software package that gives quantitative predictions for user-specified materials and experimental conditions. The method is based on modeling entire energy loss spectra, including shot noise as well as instrumental noise, and taking into account all the relevant experimental parameters. We describe the steps involved in modeling the entire spectrum, from the zero loss up to inner shell edges, and pay particular attention to the contributions to the pre-edge background. The predicted spectra are used to evaluate the signal-to-noise ratios (SNRs) for inner shell edges from user-specified minority elements. The software also predicts the minimum detectable mass (MDM) and minimum mass fraction (MMF). It can be used to ascertain whether an element present at a particular concentration should be detectable for given experimental conditions, and also to quickly and quantitatively explore ways of optimizing the experimental conditions for a particular EELS analytical task. We demonstrate the usefulness of the software by confirming the recent empirical observation of single atom detection using EELS of phosphorus in thin carbon films, and show the effect on the SNR of varying the acquisition parameters. The case of delta-doped semiconductors is also considered as an important example from materials science where low detection limits and high spatial resolution are essential, and the feasibility of such characterization using EELS is assessed.