US 11,601,509 B1Granted

Systems and methods for identifying entities between networks

AssigneeStripe, Inc.

InventorsPeter Lofgren, Pranav Dandekar, Ashish Goel, Andrew Paul Tausz

Life of this patent

Estimating the term…

Loading patent…

Abstract

Described herein are systems and methods for predicting whether an entity associated with a profile in one network is the same entity that is associated with a profile in a second network, which networks may represent networks from different network services or may represent networks from the same network service. In embodiments, network graph features, including nodes and connections, may be used to predict a probability that the profiles in the two networks should be matched. In embodiments, additional or different factors may be included in the predicted probability, such as homophily, match probabilities of seed nodes, match probabilities of attribute-matched nodes, attribute-attribute co-occurrence probabilities, and the like.

Claims20

Dependent claims

A computer-implemented method, comprising:

generating a feature vector for a query node from a first network, the feature vector including at least one feature determined based on a query image associated with the query node;
generating a reduced feature vector based on a principal component analysis (PCA) map that utilizes the feature vector as input;
selecting, using one or more k-d trees, a set of candidate feature vectors from a PCA feature database;
determining a set of candidate nodes in a second network associated with at least one candidate feature vector from the set of candidate feature vectors, the second network comprising a set of nodes, each node from the set of nodes associated with a candidate image;
for each candidate node from the set of candidate nodes, predicting a probability that assesses whether the query node is a match with the candidate node using data associated with at least one of the nodes in the first and second networks that are connected within a set number of nodes to the query node or within a set number of nodes to the candidate node, the predicting:
for a set of seed nodes comprising at least one set of matched nodes, each set of matched nodes comprising a first node from the first network and is associated with an entity and a second node from the second network and is associated with a same entity, determining a first score using connections of seed nodes to the candidate node and to the query node;
for a set of attribute-matched nodes comprising at least one set of nodes, each set of nodes comprising a node from the first network and a node from the second network that share an attribute and the nodes in the set of attribute-matched nodes are not included in the set of seed nodes, determining a second score using connections of attribute-matched nodes to the candidate node and to the query node;
and
obtaining a combined score that represents a predicted probability that an entity associated with the query node is a different entity than an entity associated with the candidate node, the combined score based at least in part on multiplying the first score with the second score;
responsive to the candidate node having a predicted probability within a threshold range, matching the query node in the first network with the candidate node in the second network;
merging, with particular data from the query node, data from the matched candidate node;
and
granting at least one network service for an account associated with the query node based at least in part on the data from the merged data.

10 more elements

A computer-implemented method, comprising:

generating a feature vector for a query node from a first network, the feature vector including at least one feature determined based on a query image associated with the query node;
generating a reduced feature vector based on a principal component analysis (PCA) map that utilizes the feature vector as input;
selecting, using one or more k-d trees, a set of candidate feature vectors from a PCA feature database;
determining a set of candidate nodes in a second network associated with at least one candidate feature vector from the set of candidate feature vectors, the second network comprising a set of nodes, each node from the set of nodes associated with a candidate image;
for each candidate node from the set of candidate nodes, predicting a probability that an entity associated with the query node is a different entity than the entity associated with the candidate node using graph features associated with the query and associated with the candidate node, the predicting comprising:
for a set of seed nodes comprising one or more sets of matched nodes, each set of matched nodes comprising a first node from a first network and is associated with an entity and a second node from the second network and is associated with the same entity, determining a first score using connections of seed nodes to the candidate node and to the query node, the first score based at least in part on multiplying a number of connections for the candidate node with a combined number of connections for the set of matched nodes in the first network and in the second network;
for a set of attribute-matched nodes comprising at least one set of nodes, each set of nodes comprising a node from the first network and a node from the second network that share an attribute and the nodes in the set of attribute-matched nodes are not included in the set of seed nodes, determining a second score using connections of attribute-matched nodes to the candidate node and to the query node;
and
obtaining a combined score that represents a predicted probability that an entity associated with the query node is a different entity than an entity associated with the candidate node, the combined score based at least in part on multiplying the first score with the second score;
and
responsive to the candidate node having a predicted probability that is below a threshold level:
assuming the entity associated with the query node in the first network is the same entity that is associated with the candidate node in the second network;
merging, with particular data from the query node, data from the matched candidate node;
and
granting at least one network service for an account associated with the query node based at least in part on the data from the merged data.

12 more elements

A system, comprising:

one or more processors;
and
a non-transitory computer-readable medium or media comprising one or more sequences of instructions which, when executed by the one or more processors, causes steps to be performed comprising:
generating a feature vector for a query node from a first network, the feature vector including at least one feature determined based on a query image associated with the query node;
generating a reduced feature vector based on a principal component analysis (PCA) map that utilizes the feature vector as input
selecting, using one or more k-d trees, a set of candidate feature vectors from a PCA feature database;
determining a set of candidate nodes in a second network associated with at least one candidate feature vector from the set of candidate feature vectors, the second network comprising a set of nodes, each node from the set of nodes associated with a candidate image;
for each candidate node from the set of candidate nodes, predicting a probability that an entity associated with the query node is a different entity than the entity associated with a candidate node using graph features associated with the query and associated with the candidate node, the predicting comprising:
for a set of seed nodes comprising at least one set of matched nodes, each set of matched nodes comprising a first node from a first network and is associated with an entity and a second node from the second network and is associated with a same entity, determining a first score using connections of seed nodes to the candidate node and to the query node, the first score based at least in part on multiplying a number of connections for the candidate node with a combined number of connections for the set of matched nodes in the first network and in the second network;
for a set of attribute-matched nodes comprising at least one set of nodes, each set of nodes comprising a node from the first network and a node from the second network that share an attribute and the nodes in the set of attribute-matched nodes are not included in the set of seed nodes, determining a second score using connections of attribute-matched nodes to the candidate node and to the query node;
and
obtaining a combined score that represents a predicted probability that an entity associated with the query node is a different entity than an entity associated with the candidate node, the combined score based at least in part on multiplying the first score with the second score;
and
responsive to the candidate node having a predicted probability that is below a threshold level:
assuming the entity associated with the query node in the first network is the same entity that is associated with the candidate node in the second network;
and
granting one or more network services to a user account associated with the query node.

14 more elements

Dependencies are inferred from the claim text.