Depending on the type of chemical system you have, you need to decide first if an EOS or a liquid activity model is most appropriate. For hydrocarbon systems, particularly if light stuff or pressure is involved, an EOS of course is natural to apply. For a chemical type system, especially if there is any dissimilarity in the species present, then the liquid activity model is best.
One excellent source of interaction parameters for either type of model is the Dechema VLE Data Series, which has quite a few volumes to it now. It covers both types of systems. More importantly, it shows you the various data sets measured and how the different models do in fitting the data, as well as how thermodynamically consistent the data points themselves are. This is important, because you should, when possible, select parameters determined from data in the region of interest to you. If you take parameters blindly from a process simulator, you don't really know the source of the data and how good the fit is. I recall at least one time when some serious modeling errors resulted from trusting the simulator too much. It turned out that the software vendor had re-fit data from multiple sources, which seemed to be those I found separately in the Dechema books, and this resulted in corrupting the fit of some good data by including some bad data. I think an automated procedure had been applied, and unintended effects resulted and were not caught by human review.