Collaborative filtering in directed graph

A computer-implemented method, at least a portion of the method being performed by a collaborative filtering computing system comprising at least at least three computing devices, the method comprising:

• accessing, by the computing system, a bipartite graph comprising a first plurality of nodes representing users and a second plurality of nodes representing a plurality of items, wherein the first plurality of nodes is interconnected with the second plurality of nodes by a plurality of edges representing a direct relationship between a specific user and a specific item;
• determining, by the computing system, more than two computing groups at least based on computing resources allocated to perform an operation, wherein the more than two computing groups are respectively associated with more than two computing devices;
• assigning, by the computing system, a plurality of sets of nodes of the first plurality of nodes to the more than two computing groups, wherein a set of nodes of the first plurality of nodes are assigned at least based on a relationship between two nodes of the first plurality of nodes;
• assigning, by the computing system, sets of multiple items from the plurality of items to the more than two computing groups;
• performing, at each of the more than two computing devices, the operation between each node of the corresponding computing group and each item of the corresponding computing group so as to form an intermediate result, wherein the intermediate result at each of the more than two computing devices comprises a rating matrix representing an interest level between each node of the corresponding computing group and each item of the corresponding computing group;
• storing, at each of the more than two computing devices, the intermediate result;
• transmitting the intermediate result from one computing device to another computing device in a cyclic order;
• updating the intermediate result by performing, at each of the more than two computing devices, the operation between each node of the corresponding computing group and each item of a different computing group transmitted by a different computing device to obtain an updated intermediate result comprising a rating matrix representing an interest level between each node of the computing group and each item of the computing group and of different computing group;
• repeating the storing step, the transmitting step, and the updating step until the operation between each item of the plurality of items and each node of the first plurality of nodes of the bipartite graph has been completed;
• generating, by the computing system, a collective result at least based on the intermediate result from each of the more than two computing devices;
• and
• saving, by the computing system, a prediction of a behavior of the plurality of users based on the collective result.

A collaborative filtering computing system, comprising:

• at least two computing devices, each computing device comprising:
• a processor;
• computer readable storage memory coupled to the processor storing computer instructions configured to:
• access a portion of a data representation of a bipartite graph comprising a first plurality of nodes representing users and a second plurality of nodes representing a plurality of items, wherein the first plurality of nodes is interconnected with the second plurality of nodes by a plurality of edges representing a relationship between a specific user and a specific item, wherein the portion of the data representation includes a selected number of nodes of the first plurality of nodes and a set of information corresponding to a set of items of the plurality of items, wherein the selected nodes are selected at least based on relationship between two nodes of the first plurality of nodes;
• perform an operation between each item of the set of items and each node of the selected number of nodes to form the set of information;
• transmit the set of information;
• receive an updated set of information, wherein the updated set of information is updated by performing the operation between each item of the set of items and an unselected group of nodes of the first plurality of nodes;
• generate a predicted behavior of each of the users associated with the selected number of nodes based on the updated set of information.