Traitement en cours

Veuillez attendre...

Paramétrages

Paramétrages

Aller à Demande

1. WO2020113208 - SYSTÈME, PROCÉDÉ ET PRODUIT-PROGRAMME D’ORDINATEUR POUR GÉNÉRER DES INTÉGRATIONS POUR DES OBJET

Note: Texte fondé sur des processus automatiques de reconnaissance optique de caractères. Seule la version PDF a une valeur juridique

[ EN ]

WHAT IS CLAIMED IS:

1. A method for generating embeddings of objects in a heterogeneous network, comprising:

receiving, with at least one processor, heterogeneous network data associated with a plurality of objects in a heterogeneous network;

selecting, with at least one processor, at least one pattern of objects; determining, with at least one processor, instances of each pattern of objects based on the heterogeneous network data;

generating, with at least one processor, a pattern matrix for each pattern of objects based on the instances of the pattern of objects;

generating, with at least one processor, pattern sequence data associated with a portion of each pattern matrix;

generating, with at least one processor, network sequence data associated with a portion of the heterogeneous network data;

combining, with at least one processor, the pattern sequence data and the network sequence data into combined sequence data; and

generating, with at least one processor, a vector for each object of the plurality of objects based on the combined sequence data.

2. The method of claim 1 , wherein the plurality of objects comprises a plurality of nodes, each node of the plurality of nodes connected to at least one other node of the plurality of nodes by an edge.

3. The method of claim 1 or 2, wherein each edge is directional.

4. The method of any of claims 1-3, wherein each node comprises a node type, and wherein the node type comprises one of a cardholder, an amount, a merchant, a merchant category code, a location, a transaction channel, a restaurant, a meal type, or a city.

5. The method of any of claims 1-4, wherein each edge comprises an edge type, the edge type comprising one of spend, pay via, at, belongs to, pay to, located in, visit, or eat.

6. The method of any of claims 1-5, wherein selecting the at least one pattern of objects comprises determining, with at least one processor, the at least one pattern of objects has a frequency of occurrence that is statistically significant based on the heterogeneous network data.

7. The method of any of claims 1 -6, wherein determining the at least one pattern of objects has the frequency of occurrence that is statistically significant comprises:

extracting, with at least one processor, the instances of the at least one pattern from the heterogeneous network data using graph submatching;

determining, with at least one processor, the frequency of occurrence of the at least one pattern of objects based on the instances of the at least one pattern of objects; and

determining, with at least one processor, the frequency of occurrence of the at least one pattern of objects is statistically significant.

8. The method of any of claims 1-7, wherein each pattern of objects comprises a motif, and wherein each pattern matrix comprises a motif adjacency matrix, and wherein generating a pattern matrix for each pattern of objects comprises:

creating, with at least one processor, a matrix for each respective motif, each row of the matrix associated with a first respective object of the plurality of objects in the heterogeneous network and each column the matrix associated with a second respective object of the plurality of objects in the heterogeneous network;

determining, with at least one processor, weights for the matrix, each weight comprising a number of the instances of the motif that includes the first respective object associated with a respective row of the matrix and the second respective object associated with a respective column of the matrix; and

storing, with at least one processor, the matrix as the motif adjacency matrix associated with the respective motif.

9. The method of any of claims 1-8, wherein the portion of each pattern matrix comprises a random walk sample of the pattern matrix; and

wherein the portion of the heterogeneous network data comprises a random walk sample of the heterogeneous network data.

10. The method of any of claims 1-9, wherein generating the vector for each object of the plurality of objects comprises:

inputting, with at least one processor, the combined sequence data into a predictive model; and

determining, with at least one processor, the vector for each object of the plurality of objects based on the output of the predictive model.

11. The method of any of claims 1-10, wherein the predictive model comprises at least one of a neural network or a skip-gram model.

12. A system for generating embeddings of objects in a heterogeneous network, the system comprising:

at least one processor programmed or configured to:

receive heterogeneous network data associated with a plurality of objects in a heterogeneous network;

determine that at least one pattern of objects included in the heterogeneous network has a frequency of occurrence that is statistically significant;

select at least one pattern of objects based on determining that the at least one pattern of objects has a frequency of occurrence that is statistically significant;

determine instances of each pattern of objects based on the heterogeneous network data;

generate a pattern matrix for each pattern of objects based on the instances of the pattern of objects;

generate pattern sequence data associated with a portion of each pattern matrix;

generate network sequence data associated with a portion of the heterogeneous network data;

combine the pattern sequence data and the network sequence data into combined sequence data; and

generate a vector for each object of the plurality of objects based on the combined sequence data.

13. The system of claim 12, wherein the plurality of objects comprises a plurality of nodes, each node of the plurality of nodes connected to at least one other node of the plurality of nodes by an edge.

14. The system of claim 12 or 13, wherein when determining the at least one pattern of objects included in the heterogeneous network has the frequency of occurrence that is statistically significant, the at least one processor is programmed or configured to:

extract the instances of the at least one pattern from the heterogeneous network data using graph submatching;

determine the frequency of occurrence of the at least one pattern of objects based on the instances of the at least one pattern of objects; and

determine the frequency of occurrence of the at least one pattern of objects is statistically significant.

15. The system of any of claims 12-14, wherein each pattern of objects comprises a motif, and wherein each pattern matrix comprises a motif adjacency matrix, and wherein, when generating a pattern matrix for each pattern of objects, the at least one processor is programmed or configured to:

create a matrix for each respective motif, each row of the matrix associated with a first respective object of the plurality of objects in the heterogeneous network and each column the matrix associated with a second respective object of the plurality of objects in the heterogeneous network;

determine weights for the matrix, each weight comprising a number of the instances of the motif that includes the first respective object associated with a respective row of the matrix and the second respective object associated with a respective column of the matrix; and

store the matrix as the motif adjacency matrix associated with the respective motif.

16. A method for generating embeddings of objects in a heterogeneous network, comprising:

receiving, with at least one processor, heterogeneous network data associated with a plurality of objects in a heterogeneous network;

selecting, with at least one processor, at least one pattern of objects;

determining, with at least one processor, instances of each pattern of objects based on the heterogeneous network data;

generating, with at least one processor, a pattern matrix for each pattern of objects based on the instances of the pattern of objects;

generating, with at least one processor, pattern sequence data associated with a portion of each pattern matrix;

generating, with at least one processor, network sequence data associated with a portion of the heterogeneous network data;

combining, with at least one processor, the pattern sequence data and the network sequence data into combined sequence data; and

generating, with at least one processor, a vector for each object of the plurality of objects based on the combined sequence data.

17. The method of claim 16, wherein the plurality of objects comprises a plurality of nodes, each node of the plurality of nodes connected to at least one other node of the plurality of nodes by an edge.

18. The method of claim 17, wherein each edge is directional.

19. The method of claim 17, wherein each node comprises a node type, and wherein the node type comprises one of a cardholder, an amount, a merchant, a merchant category code, a location, a transaction channel, a restaurant, a meal type, or a city.

20. The method of claim 19, wherein each edge comprises an edge type, the edge type comprising one of spend, pay via, at, belongs to, pay to, located in, visit, or eat.

21. The method of claim 16, wherein selecting the at least one pattern of objects comprises determining, with at least one processor, the at least one pattern of objects has a frequency of occurrence that is statistically significant based on the heterogeneous network data.

22. The method of claim 21 , wherein determining the at least one pattern of objects has the frequency of occurrence that is statistically significant comprises:

extracting, with at least one processor, the instances of the at least one pattern from the heterogeneous network data using graph submatching;

determining, with at least one processor, the frequency of occurrence of the at least one pattern of objects based on the instances of the at least one pattern of objects; and

determining, with at least one processor, the frequency of occurrence of the at least one pattern of objects is statistically significant.

23. The method of claim 16, wherein each pattern of objects comprises a motif, and wherein each pattern matrix comprises a motif adjacency matrix, and wherein generating a pattern matrix for each pattern of objects comprises:

creating, with at least one processor, a matrix for each respective motif, each row of the matrix associated with a first respective object of the plurality of objects in the heterogeneous network and each column the matrix associated with a second respective object of the plurality of objects in the heterogeneous network;

determining, with at least one processor, weights for the matrix, each weight comprising a number of the instances of the motif that includes the first respective object associated with a respective row of the matrix and the second respective object associated with a respective column of the matrix; and

storing, with at least one processor, the matrix as the motif adjacency matrix associated with the respective motif.

24. The method of claim 16, wherein the portion of each pattern matrix comprises a random walk sample of the pattern matrix; and

wherein the portion of the heterogeneous network data comprises a random walk sample of the heterogeneous network data.

25. The method of claim 16, wherein generating the vector for each object of the plurality of objects comprises:

inputting, with at least one processor, the combined sequence data into a predictive model; and

determining, with at least one processor, the vector for each object of the plurality of objects based on the output of the predictive model.

26. The method of claim 25, wherein the predictive model comprises at least one of a neural network or a skip-gram model.

27. A system for generating embeddings of objects in a heterogeneous network, the system comprising:

at least one processor programmed or configured to:

receive heterogeneous network data associated with a plurality of objects in a heterogeneous network;

determine that at least one pattern of objects included in the heterogeneous network has a frequency of occurrence that is statistically significant;

select at least one pattern of objects based on determining that the at least one pattern of objects has a frequency of occurrence that is statistically significant;

determine instances of each pattern of objects based on the heterogeneous network data;

generate a pattern matrix for each pattern of objects based on the instances of the pattern of objects;

generate pattern sequence data associated with a portion of each pattern matrix;

generate network sequence data associated with a portion of the heterogeneous network data;

combine the pattern sequence data and the network sequence data into combined sequence data; and

generate a vector for each object of the plurality of objects based on the combined sequence data.

28. The system of claim 27, wherein the plurality of objects comprises a plurality of nodes, each node of the plurality of nodes connected to at least one other node of the plurality of nodes by an edge.

29. The system of claim 27, wherein when determining the at least one pattern of objects included in the heterogeneous network has the frequency of occurrence that is statistically significant, the at least one processor is programmed or configured to:

extract the instances of the at least one pattern from the heterogeneous network data using graph submatching;

determine the frequency of occurrence of the at least one pattern of objects based on the instances of the at least one pattern of objects; and

determine the frequency of occurrence of the at least one pattern of objects is statistically significant.

30. The system of claim 27, wherein each pattern of objects comprises a motif, and wherein each pattern matrix comprises a motif adjacency matrix, and wherein, when generating a pattern matrix for each pattern of objects, the at least one processor is programmed or configured to:

create a matrix for each respective motif, each row of the matrix associated with a first respective object of the plurality of objects in the heterogeneous network and each column the matrix associated with a second respective object of the plurality of objects in the heterogeneous network;

determine weights for the matrix, each weight comprising a number of the instances of the motif that includes the first respective object associated with a respective row of the matrix and the second respective object associated with a respective column of the matrix; and

store the matrix as the motif adjacency matrix associated with the respective motif.

31. A computer program product for generating embeddings of objects in a heterogeneous network, comprising at least one non-transitory computer-readable medium comprising one or more instructions that, when executed by at least one processor, cause the at least one processor to:

receive heterogeneous network data associated with a plurality of objects in a heterogeneous network;

determine that at least one pattern of objects included in the heterogeneous network has a frequency of occurrence that is statistically significant; select at least one pattern of objects based on determining that the at least one pattern of objects has a frequency of occurrence that is statistically significant;

determine instances of each pattern of objects based on the heterogeneous network data;

generate a pattern matrix for each pattern of objects based on the instances of the pattern of objects;

generate pattern sequence data associated with a portion of each pattern matrix;

generate network sequence data associated with a portion of the heterogeneous network data;

combine the pattern sequence data and the network sequence data into combined sequence data; and

generate a vector for each object of the plurality of objects based on the combined sequence data.

32. The computer program product of claim 31 , wherein the plurality of objects comprises a plurality of nodes, each node of the plurality of nodes connected to at least one other node of the plurality of nodes by an edge.

33. The computer program product of claim 31 , wherein the one or more instructions that cause the at least one processor to determine that the at least one pattern of objects included in the heterogeneous network has the frequency of occurrence that is statistically significant cause the at least one processor to:

extract the instances of the at least one pattern from the heterogeneous network data using graph submatching;

determine the frequency of occurrence of the at least one pattern of objects based on the instances of the at least one pattern of objects; and

determine the frequency of occurrence of the at least one pattern of objects is statistically significant.

34. The computer program product of claim 31 , wherein each pattern of objects comprises a motif, and wherein each pattern matrix comprises a motif adjacency matrix, and wherein, when generating a pattern matrix for each pattern of objects, wherein the one or more instructions further cause the at least one processor to:

create a matrix for each respective motif, each row of the matrix associated with a first respective object of the plurality of objects in the heterogeneous network and each column the matrix associated with a second respective object of the plurality of objects in the heterogeneous network;

determine weights for the matrix, each weight comprising a number of the instances of the motif that includes the first respective object associated with a respective row of the matrix and the second respective object associated with a respective column of the matrix; and

store the matrix as the motif adjacency matrix associated with the respective motif.

35. The computer program product of claim 31 , wherein the one or more instructions that cause the at least one processor to generate the vector for each object of the plurality of objects cause the at least one processor to:

input the combined sequence data into a predictive model; and determine the vector for each object of the plurality of objects based on the output of the predictive model.