I don't think the predictive power of models build from data driven happiness research is very high. I wouldn't ignore the research completely but there nothing rational about using a model just because it's data based if nobody showed that the model is useful for prediction in the relevant domain.