How to recommend apps to a user based on functionality, permissions, and user interests is the subject of this paper. Traditional app recommendation algorithms are based on popularity and downloads. However, these algorithms usually do not take into consideration users’ privacy concerns. For example, users tend to pay close attention to access permissions requested by apps due to privacy and security concerns. Few existing recommendation algorithms take a balanced approach that addresses all three issues of functionality, permissions, and user interests. Some only consider permissions while neglecting user interests, while others simply perform a linear combination of all three factors, which is not as effective.

The authors propose and validate “a novel matrix factorization algorithm MFPF” based on both permissions and user interests. The basic algorithm models the relation between user interest, functionality, and permissions as the problem of finding the best solution in a latent matrix of the three involved factors. A control value (α) in the solution can be adjusted to reach best performance.

The algorithm was tested using a set of real data from the Anzhi Market. The cleansed data included 1287 apps, 975 users, and 98621 comments. The authors used 80 percent of the raw data as their training data and the remaining 20 percent as testing data. Two criteria were tested: the root-mean-square error (RMSE) and the precision of top-N recommendations. The authors then compared these measures to some other leading algorithms in two separate groups. The first group consisted of three traditional recommendation algorithms, and the second group focused on the app’s functionality and permissions. The proposed MFPF algorithm outperformed all other algorithms in recommendation accuracy.

The proposed algorithm addresses an important issue in app recommendation, that is, balancing the factors of functionality, access rights requested by apps, and user interests. It is proven to be an effective tool.