Systems and methods for determining application installation likelihood based on probabilistic combination of subordinate methods

摘要

A method for providing a scored list of applications for use in recommending applications to a user includes receiving at a cloud computing service a request to provide a composite scored list of applications for a user, where the cloud computing service provides a set of available applications to each user and receives input data for determining the composite scored list of applications. The cloud computing service initializes a plurality of independent scorers, which each take as input a portion of the input data and outputs a scored list of applications, and normalizes the scored list of applications outputted by each independent scorer. The cloud computing service constructs the composite scored list of applications, where the composite scored list of applications is a function of the normalized scored list of applications and a weighting value assigned to each independent scorer, and outputs the composite scored list of applications.

主权项

1. A method for providing recommendations for applications to a user, the method comprising:
receiving at a cloud computing service a request to provide a composite scored list of applications for a user, wherein the cloud computing service provides a predetermined set of available applications to each user of the cloud computing service; receiving input data for determining the composite scored list of applications; initializing a plurality of independent scorers, wherein each independent scorer takes as input a portion of the input data and outputs a scored list of applications in the predetermined set of available applications; normalizing the scored list of applications outputted by each independent scorer; constructing the composite scored list of applications, wherein the composite scored list of applications is a function of the normalized scored list of applications outputted by each independent scorer and a weighting value assigned to each independent scorer, and wherein a composite score for a first application in the predetermined set of available applications is calculated as:
(IS1*W1+IS2*W2+ . . . +ISN*WN)/(W1+W2+ . . . +WN)where N is a number of independent scorers, ISn is a normalized score of the first application from the n-th independent scorer, and Wn is the weighting value assigned to the n-th independent scorer; and
outputting the composite scored list of applications.