Privacy preserving user group expansion
摘要
說明書
In some aspects, the method can include generating, for each user in the third set of users, a score based on a set of data including (i) the user identifier associated with the user, (ii) the user interest group, (iii) an entity group of the entity, (iv) electronic resources of the entity, and (v) keywords associated with the entity; and selecting, for inclusion in the second set of users, each user having a score that satisfies a threshold score condition for the user list. In some aspects, generating, for each user in the third set of users, a score can include receiving the set of data; providing the set of data as input to a machine learning model that was trained to correlate training sets of data with likelihood of a user performing one or more specified actions to determine first and second arrays, wherein the first array corresponds to a user embedding associated with the user and the second array corresponds to an entity embedding associated with the entity; determining a distance between the first array and the second array; and assigning the score for the user based on the determined distance between the first array and the second array. In some aspects, the method can also include ranking the third set of users in the intermediate list based on the assigned score for each user in the third set of users; and selecting, for inclusion in the second set of users, each user from the ranked intermediate list having the assigned score that exceeds the threshold score condition for the user list. In yet some aspects, the machine learning model is a deep neural network (DNN).
