Event prediction
Computer-implemented prediction methods, in accordance with various embodiments, involve storing key-value pairs each including a key tuple and an associated value in a data store, building a prediction model associated with at least a subset of the key-value pairs, and in response to receipt of a prediction query that specifies a key tuple, generating a prediction response based on the prediction model that includes a predicted value for the specified key tuple. In one example application, the key tuples each include an operation specifying a purchase and a two-part identifier specifying a customer and a product.
1. A prediction method implemented by a computer, the method comprising:
storing, in a data store implemented in one or more non-transitory machine-readable media, a plurality of key-value pairs each comprising a key tuple and an associated value, the key tuple specifying an operation and an identifier;
building, with a computer processor, a prediction model associated with at least a subset of the plurality of key-value pairs;
receiving a prediction query specifying a key tuple comprising an operation specifying a purchase and a two-part identifier specifying a first customer and a first product, wherein the plurality of key-value pairs does not include a key-value pair including a value for the key tuple specified in the prediction query; and
generating, with the computer processor, a prediction response based on the prediction model, the prediction response comprising a predicted value for the key tuple specified in the prediction query and being based on key-value pairs in the data store that have keys each comprising an operation specifying the purchase and a two-part identifier specifying the first customer and a product different from the first product or specifying a customer different from the first customer and the first product.
2. The method of claim 1 , wherein:
the plurality of key-value pairs includes one or more key-value pairs for the specified operation and one or more key-value pairs for the specified identifier.
3. The method of claim 1 , further comprising:
ingesting new data from one or more data sources;
converting the new data into new key-value pairs for storage in the data store; and
updating the prediction model based on the new key-value pairs.
4. The method of claim 1 , wherein the prediction model is built using a model definition received from a user.
5. The method of claim 1 , wherein the prediction model is a first prediction model associated with a first predictive task, the method further comprising building one or more additional prediction models based on respective subsets of the plurality of key-value pairs.
6. The method of claim 5 , wherein the first prediction model and the one or more additional prediction models are built for different respective predictive tasks.
7. The method of claim 5 , wherein the first prediction model and the one or more additional prediction models are built for a same prediction task, the method further comprising performing cross-validation techniques to assess relative performances of the prediction model and the one or more additional prediction models on the prediction task.
8. The method of claim 1 , wherein the plurality of key-value pairs have associated confidence values that decrease with time.
9. The method of claim 1 , wherein the at least a subset of the plurality of key-value pairs with which the prediction model is associated results from filtering the plurality of key-value pairs based on user input.
10. The method of claim 1 , wherein the key tuples of the plurality of key-value pairs collectively comprise multiple operations, and wherein the prediction model is built based on a subset of the multiple operations specified by a user.
11. The method of claim 1 , wherein the prediction response is related to a future transaction.
12. A prediction method implemented by a computer, the method comprising:
storing, in a data store implemented in one or more non-transitory machine-readable media, a plurality of key-value pairs each comprising a key tuple and an associated value, the key tuple specifying an operation and an identifier;
building, with a computer processor, a prediction model associated with at least a subset of the plurality of key-value pairs;
receiving a prediction query specifying a key tuple comprising a specified operation and a specified identifier, wherein the plurality of key-value pairs does not include a key-value pair including a value for the key tuple specified in the prediction query; and
generating, with the computer processor, a prediction response based on the prediction model, the prediction response comprising a predicted value for the key tuple specified in the prediction query, wherein:
the specified operation of the prediction query is a first operation and the specified identifier of the prediction query is a first identifier;
the plurality of key-value pairs in the data store comprises: a first key-value pair with a key specifying a second operation and the first identifier, a second key-value pair with a key specifying the second operation and a second identifier, and a third key-value pair with a key specifying the first operation and the second identifier; and
the prediction response is based on similarity between values of the first and second key-value pairs in conjunction with a value of the third key-value pair.
13. The method of claim 12 , wherein the plurality of key-value pairs represent patient data, and the prediction response comprises a medical treatment recommendation.
14. A computer system comprising:
a data store, implemented in one or more non-transitory machine-readable media, storing a plurality of key-value pairs each comprising a key tuple and an associated value, the key tuple specifying an operation and an identifier;
a model definition model, implemented using one or more computer processors, configured to build a prediction model from at least a subset of the plurality of key-value pairs; and
a prediction engine, implemented using the one or more computer processors, configured to generate a prediction response based on the prediction model in response to a prediction query that specifies a key tuple comprising a specified operation and a specified identifier,
wherein:
the plurality of key-value pairs does not include a key-value pair including a value for the key tuple specified in the prediction query;
the prediction response comprises a predicted value for the key tuple specified in the prediction query;
the specified operation of the prediction query is a first operation and the specified identifier of the prediction query is a first identifier;
the plurality of key-value pairs in the data store comprises: a first key-value pair with a key specifying a second operation and the first identifier, a second key-value pair with a key specifying the second operation and a second identifier, and a third key-value pair with a key specifying the first operation and the second identifier; and
the prediction response is based on similarity between values of the first and second key-value pairs in conjunction with a value of the third key-value pair.
15. The system of claim 14 , wherein the data store is in operable communication with multiple data sources via associated data connectors.
16. The system of claim 15 , wherein the data connectors are configured to enable ingestion, into the data store, of multiple of: streaming data, bulk data from a flat file, or data from a database.
17. One or more non-transitory computer-readable storage media storing:
a data store comprising a plurality of key-value pairs each comprising a key tuple and an associated value, the key tuple specifying an operation and an identifier; and
processor-executable instructions configured to cause a processor to perform operations comprising:
building a prediction model associated with at least a subset of the plurality of key-value pairs, and
in response to receipt of a prediction query specifying a key tuple comprising an operation specifying a purchase and a two-part identifier specifying a first customer and a first product, generating a prediction response based on the prediction model, the prediction response comprising a predicted value for the key tuple specified in the prediction query and being based on key-value pairs in the data store that have keys each comprising an operation specifying the purchase and a two-part identifier specifying the first customer and a product different from the first product or specifying a customer different from the first customer and the first product,
wherein the plurality of key-value pairs does not include a key-value pair including a value for the key tuple specified in the prediction query.