Predicting entity outcomes using taxonomy classifications of transactions
Methods, systems, and computer programs, for predicting a likely outcome for an entity. A method includes obtaining a first data structure that includes data that represents a transaction, determining a similarity level of the transaction to each of a plurality of categories, determining a transaction category based on the determined similarity level, generating an input data structure that includes data representing (i) at least a portion of the data representing the transaction, and (ii) data describing the determined category, providing the input data structure to a predictive algorithm trained to determine a value that represents a likely outcome for an entity that initiated a transaction, obtaining output generated by the predictive algorithm based on the predictive algorithm's processing of the input data structure, the output including a value that represents a likely outcome for the entity, and determining a likely outcome for the entity based on the obtained output.
1. A system for verifying point-of-sale transactions, the system comprising:
one or more processors; and
one or more computer storage medium storing instructions that, when executed by the one or more processors, causes the one or more processors to perform operations, the operations comprising:
obtaining, by the one or more processors, one or more first data structures that each include fields structuring data that represents a particular point-of-sale transaction initiated by an entity;
storing the one or more obtained first data structures in a solid state memory device; and
for each particular first data structure of the one or more first data structures stored in the solid state memory device:
determining, by the one or more processors, a level of similarity of the transaction represented by the particular first data structure to each of a plurality of predetermined categories;
determining, by the one or more processors, a category of the transaction represented by the particular first data structure based on the determined level of similarity;
generating, by the one or more processors, an input data structure that includes fields structuring data representing (i) at least a portion of the data representing the transaction that is structured by the first data structure and (ii) data describing the determined category;
providing, by the one or more processors, the input data structure to a predictive algorithm that has been trained to determine a value that represents a probability that a point-of-sale transaction initiated by an entity is fraudulent;
obtaining, by the one or more processors, output data generated by the predictive algorithm based on the predictive algorithm's processing of the input data structure, wherein the output data includes a value that represents a probability that the particular point-of-sale transaction initiated by the entity and represented by the particular first data structure is fraudulent; and
determining, by the one or more processors, whether the particular point-of-sale transaction initiated by the entity and represented by the first data structure is fraudulent based on the obtained output data.
2. The system of claim 1 , wherein the operations further comprise:
for each particular first data structure of the one or more first data structures:
determining, by the one or more processors, that a relationship exists between the data representing the transaction that is structured by the particular first data structure and data representing one or more other transactions; and
generating, by the one or more processors, one or more second data structures representing (i) at least a portion of the data representing the financial transaction that is structured by the particular first data structure and (ii) data describing the determined relationship of the financial transaction that is structured by the particular first data structure to one or more other transactions represented by first data structures;
wherein determining, by the one or more processors, a level of similarity of the transaction represented by the particular first data structure to each of a plurality of predetermined categories comprises:
determining, by the one or more processors, a level of similarity of the transaction represented by the particular second data structure to each of a plurality of predetermined categories;
wherein determining, by the one or more processors, a category of the transaction represented by the particular first data structure based on the determined level of similarities comprises:
determining, by the one or more processors, a category of the transaction represented by the particular second data structure based on the determined level of similarities;
wherein generating, by the one or more processors, an input data structure that includes fields structuring data representing (i) at least a portion of the data representing the transaction that is structured by the particular first data structure and (ii) data describing the determined category comprises:
generating, by the one or more processors, an input data structure that includes fields structuring data representing (i) at least a portion of the data representing the transaction that is structured by the second data structure and (ii) data describing the determined category.
3. The system of claim 1 , wherein the generated input data structure can also include one or more fields structuring data that represents non-financial information.
4. A system for preventing financial loss by a financial institution, the system comprising:
one or more processors; and
one or more computer storage medium storing instructions that, when executed by the one or more processors, causes the one or more processors to perform operations, the operations comprising:
obtaining, by the one or more processors, one or more first data structures that each include fields structuring data that represents a transaction, wherein the transaction includes a request for a particular loan by an entity;
storing the one or more obtained first data structures in a solid state memory device; and
for each particular first data structure of the one or more first data structures stored in the solid state memory device:
determining, by the one or more processors, a level of similarity of the transaction represented by the particular first data structure to each of a plurality of predetermined categories;
determining, by the one or more processors, a category of the transaction represented by the particular first data structure based on the determined level of similarity;
generating, by the one or more processors, an input data structure that includes fields structuring data representing (i) at least a portion of the data representing the transaction that is structured by the first data structure and (ii) data describing the determined category;
providing, by the one or more processors, the input data structure to a predictive algorithm that has been trained to determine a value that represents a probability of default on a loan by an entity;
obtaining, by the one or more processors, output data generated by the predictive algorithm based on the predictive algorithm's processing of the input data structure, wherein the output data includes a value that represents a probability that the entity would default on the particular loan that is requested by the entity and represented by the particular first data structure; and
determining, by the one or more processors, whether the entity is likely to default on the particular loan that is requested by the entity and represented by the particular first data structure based on the obtained output data.
5. The system of claim 4 , wherein obtaining, by the one or more processors, one or more first data structures that include fields structuring data that represents a transaction comprises:
accessing, by the one or more processors, a database of a financial institution that stores multiple first data structures that each include fields structuring data that represents a particular financial transaction that was previously initiated by the entity.
6. The system of claim 1 , wherein obtaining, by the one or more processors, one or more first data structures that include fields structuring data that represents a transaction comprises:
accessing, by the one or more processors, data representing a real-time transaction before the real-time transaction is authorized, the data representing the real-time transaction comprising at least one first data structure that includes fields structuring data that represents the real-time transaction.
7. The system of claim 1 , wherein the category of the transaction is a transaction classification of a taxonomy classifier.
8. A method for preventing financial harm, the method comprising:
obtaining, by the one or more processors, one or more first data structures that each include fields structuring data that represents a transaction, wherein the transaction includes (i) a particular point-of-sale transaction initiated by an entity or (ii) a request for a particular loan by an entity;
storing the one or more obtained first data structures in a solid state memory device; and
for each particular first data structure of the one or more first data structures stored in the solid state memory device:
determining, by the one or more processors, a level of similarity of the transaction represented by the particular first data structure to each of a plurality of predetermined categories;
determining, by the one or more processors, a category of the transaction represented by the particular first data structure based on the determined level of similarity;
generating, by the one or more processors, an input data structure that includes fields structuring data representing (i) at least a portion of the data representing the transaction that is structured by the first data structure and (ii) data describing the determined category;
providing, by the one or more processors, the input data structure to a predictive algorithm that has been trained to determine (i) a value that represents a probability that a point-of-sale transaction initiated by an entity is fraudulent or (ii) a value that represents a probability of default on a loan by an entity;
obtaining, by the one or more processors, output data generated by the predictive algorithm based on the predictive algorithm's processing of the input data structure, wherein the output data includes (i) a value that represents a probability that the particular point-of-sale transaction initiated by the entity and represented by the particular first data structure is fraudulent or (ii) a value that represents a probability that the entity would default on the particular loan that is requested by the entity and represented by the particular first data structure; and
determining, by the one or more processors, (i) whether the particular point-of-sale transaction initiated by the entity and represented by the first data structure is fraudulent based on the obtained output data or (ii) whether the entity is likely to default on the particular loan that is requested by the entity and represented by the particular first data structure based on the obtained output data.
9. The method of claim 8 , wherein obtaining, by the one or more processors, one or more first data structures that include fields structuring data that represents a transaction comprises:
accessing, by the one or more processors, a database of a financial institution that stores multiple first data structures that each include fields structuring data that represents a particular financial transaction that was previously initiated by the entity.
10. The method of claim 8 , obtaining, by the one or more processors, one or more first data structures that include fields structuring data that represents a transaction comprises:
accessing, by the one or more processors, data representing a real-time transaction before the real-time transaction is authorized, the data representing the real-time transaction comprising at least one first data structure that includes fields structuring data that represents the real-time transaction.
11. The method of claim 1 , the method further comprising:
for each particular first data structure of the one or more first data structures:
determining, by the one or more processors, that a relationship exists between the data representing the transaction that is structured by the particular first data structure and data representing one or more other transactions; and
generating, by the one or more processors, one or more second data structures representing (i) at least a portion of the data representing the financial transaction that is structured by the particular first data structure and (ii) data describing the determined relationship of the financial transaction that is structured by the particular first data structure to one or more other transactions represented by first data structures;
wherein determining, by the one or more processors, a level of similarity of the transaction represented by the particular first data structure to each of a plurality of predetermined categories comprises:
determining, by the one or more processors, a level of similarity of the transaction represented by the particular second data structure to each of a plurality of predetermined categories;
wherein determining, by the one or more processors, a category of the transaction represented by the particular first data structure based on the determined level of similarities comprises:
determining, by the one or more processors, a category of the transaction represented by the particular second data structure based on the determined level of similarities;
wherein generating, by the one or more processors, an input data structure that includes fields structuring data representing (i) at least a portion of the data representing the transaction that is structured by the particular first data structure and (ii) data describing the determined category comprises:
generating, by the one or more processors, an input data structure that includes fields structuring data representing (i) at least a portion of the data representing the transaction that is structured by the second data structure and (ii) data describing the determined category.
12. A non-transitory computer-readable medium storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to perform operations comprising:
obtaining, by the one or more processors, one or more first data structures that each include fields structuring data that represents a transaction, wherein the transaction includes (i) a particular point-of-sale transaction initiated by an entity or (ii) a request for a particular loan by an entity;
storing the one or more obtained first data structures in a solid state memory device; and
for each particular first data structure of the one or more first data structures stored in the solid state memory device:
determining, by the one or more processors, a level of similarity of the transaction represented by the particular first data structure to each of a plurality of predetermined categories;
determining, by the one or more processors, a category of the transaction represented by the particular first data structure based on the determined level of similarity;
generating, by the one or more processors, an input data structure that includes fields structuring data representing (i) at least a portion of the data representing the transaction that is structured by the first data structure and (ii) data describing the determined category;
providing, by the one or more processors, the input data structure to a predictive algorithm that has been trained to determine (i) a value that represents a probability that a point-of-sale transaction initiated by an entity is fraudulent or (ii) a value that represents a probability of default on a loan by an entity;
obtaining, by the one or more processors, output data generated by the predictive algorithm based on the predictive algorithm's processing of the input data structure, wherein the output data includes (i) a value that represents a probability that the particular point-of-sale transaction initiated by the entity and represented by the particular first data structure is fraudulent or (ii) a value that represents a probability that the entity would default on the particular loan that is requested by the entity and represented by the particular first data structure; and
determining, by the one or more processors, (i) whether the particular point-of-sale transaction initiated by the entity and represented by the first data structure is fraudulent based on the obtained output data or (ii) whether the entity is likely to default on the particular loan that is requested by the entity and represented by the particular first data structure based on the obtained output data.