Passively collected mobile data, particularly large-scale datasets, have become increasingly popular in transportation studies for understanding human mobility patterns and evaluating transportation system performance. However, missing data issues are commonly encountered and often overlooked or insufficiently addressed. This paper explores the application of recommendation system (RecSys) models to infer missing data in passively collected mobility datasets. We first examine the similarities and differences between RecSys and transportation mobility data, which leads us to adapt RecSys models using non-negative matrix factorization (NMF) for estimating missing mobility data. We conduct numerical experiments with app-based mobility data from the Seattle area to evaluate the proposed method.