Method for providing data science, artificial intelligence and machine learning as-a-service

Abstract

An automated method of predictive model development first cleans up raw supervised and unsupervised training data with a step that uses an algorithm to make every field of every record consistent, cohesive, and productive. Then the resulting flat data is given texture in a next step by a data enrichment algorithm that culls fields that do not contribute to predictive model building and that adds new fields computed from data combinations that are tested to add value to later steps that build different types of predictive models. Another late step for building smart-agents and their entity profiles uses another algorithm that benefits greatly from the cleaned and highly enriched training data. The predictive models and smart-agents and their entity profiles are then rendered as deliverable predictive model markup language documents in a final step executed by a specialized algorithm.

1 . A method that provides artificial intelligence and machine learning as-a-service by generating predictive models from training data input records, comprising:
 receiving with a network input, electronic records of supervised and/or unsupervised training data of a user-service consumer that represent a plurality of records having a predefined plurality of data fields, and structured and unstructured text, and decrypting either or both with an apparatus for executing a decoding algorithm, and transforming a series of results into a set of non-transitory raw-data records that are collectively stored in a machine-readable storage mechanism;   cleaning and improving the integrity of the data stored in the raw-data records with an apparatus for executing a data integrity analysis algorithm that compares and corrects data values in each data field, and that deduces the context of any included structured and unstructured text with an apparatus for executing a contextual dictionary algorithm, and transforming a result into a set of flat-data records that are collectively stored in a machine-readable storage mechanism;   if the number of records in the set of flat-data records exceeds a predefined threshold data value stored in the machine-readable storage mechanism, then sampling a portion of the set of flat-data records and parsing the samples for storage into a training set, a test set, and a blind set by a machine-readable storage mechanism, and accessible thereafter for use in the remaining steps, all with an apparatus for executing a sampling algorithm that limits the number of records that must be processed by the remaining steps;   enhancing, enriching, and concentrating the sampled or unsampled data stored in the flat-data records with an apparatus for executing a data enrichment algorithm that removes data in particular data fields and the data fields themselves that are empirically predetermined to contribute little to building a predictive model, and that calculates and combines any data into new data fields that are predetermined to contribute most to building said predictive model, and that converts any text into a context classification with an apparatus for executing a context mining algorithm, and then transforming a plurality of intermediate results into a set of enriched-data records that are collectively stored in a machine-readable storage mechanism;   building a plurality of smart-agent predictive models for each entity identifiable in the samples of the enriched-data records with an apparatus for executing a smart-agent building algorithm;   selecting amongst the plurality of smart-agent predictive models obtained in the building and updating a corresponding particular smart-agent's real-time profile and long-term profile stored in a machine-readable storage mechanism with data from samples of the enriched-data records that correspond to a transaction activity of a particular entity with an apparatus for executing a smart-agent algorithm that compares a current transaction to previously memorialized transactions, and transforming and storing a result of that as smart-agent IFM-type XML document in a machine-readable storage mechanism;   exporting through a network output to said user-service consumer at least one of the predictive model markup language documents with an apparatus for executing a data-science-as-a-service algorithm from a network server.   
     
     
         2 . The method of  claim 1  further comprising:
 building a data mining predictive model by applying the same data from the samples of the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a data mining algorithm, and transforming a result into a data-mining predictive model markup language document that is stored in a machine-readable storage mechanism. 
 
     
     
         3 . The method of  claim 1  further comprising:
 building a neural network predictive model by applying the same data from the samples of the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a neural network algorithm, and transforming a result into a neural-network predictive model markup language document that is stored in a machine-readable storage mechanism. 
 
     
     
         4 . The method of  claim 1  further comprising:
 building a case-based-reasoning predictive model by applying the same data from the samples of the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a cased-based reasoning algorithm, and transforming a result into a case-based-reasoning predictive model markup language document that is stored in a machine-readable storage mechanism. 
 
     
     
         5 . The method of  claim 1  further comprising:
 building a clustering predictive model by applying the same data from the samples of the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a clustering algorithm, and transforming a result into a clustering predictive model markup language document that is stored in a machine-readable storage mechanism. 
 
     
     
         6 . The method of  claim 1  further comprising:
 building a business rules predictive model by applying the same data from the samples of the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a business rules algorithm, and transforming a result into a business rules predictive model markup language document that is stored in a machine-readable storage mechanism. 
 
     
     
         7 . The method of  claim 1  further comprising:
 building a data mining predictive model capable of scoring a record and assigning a confidence to a corresponding score by applying the same data from the samples of the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a data mining algorithm, and transforming a result into a data-mining predictive model markup language document that is stored in a machine-readable storage mechanism; 
 building a neural network predictive model capable of scoring a record and assigning a confidence to a corresponding score by applying the same data from the samples of the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a neural network algorithm, and transforming a result into a neural-network predictive model markup language document that is stored in a machine-readable storage mechanism; 
 building a case-based-reasoning predictive model capable of scoring a record and assigning a confidence to a corresponding score by applying the same data from the samples of the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a cased-based reasoning algorithm, and transforming a result into a case-based-reasoning predictive model markup language document that is stored in a machine-readable storage mechanism; 
 building a clustering predictive model capable of scoring a record and assigning a confidence to a corresponding score by applying the same data from the samples of the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a clustering algorithm, and transforming a result into a clustering predictive model markup language document that is stored in a machine-readable storage mechanism; and 
 building a business rules predictive model capable of scoring a record and assigning a confidence to a corresponding score by applying the same data from the samples of the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a business rules algorithm, and transforming a result into a business rules predictive model markup language document that is stored in a machine-readable storage mechanism. 
 
     
     
         8 . The method of  claim 7  further comprising:
 working at least two of the predictive models with an apparatus for executing an algorithm to produce a score and a score-confidence level in parallel sets from a particular record in a plurality of enriched-data records, and combining a result with an apparatus for executing an algorithm in a computer network server to form a decision that selects a winning predictive model output based on one of a business rule, a highest confidence winner-take-all, and a confidence-weighted blend, and storing a decision data in a machine-readable storage mechanism. 
 
     
     
         9 . A method of prediction that assists managers in making business decisions from information provided by electronic data records, comprising:
 generating a plurality of dissimilar predictive models in preparation for parallel operation together with a single decision output derived from several predictions, the generating accomplished by an apparatus for executing an algorithm for:
 receiving with a network input, electronic records of supervised and/or unsupervised training data of a user-service consumer that represent a plurality of records having a predefined plurality of data fields, and structured and unstructured text, and decrypting either or both with an apparatus for executing a decoding algorithm, and transforming a series of results into a set of non-transitory raw-data records that are collectively stored in a machine-readable storage mechanism; 
 cleaning and improving the integrity of the data stored in the raw-data records with an apparatus for executing a data integrity analysis algorithm that compares and corrects data values in each data field, and that deduces the context of any included structured and unstructured text with an apparatus for executing a contextual dictionary algorithm, and transforming a result into a set of flat-data records that are collectively stored in a machine-readable storage mechanism; 
 if the number of records in the set of flat-data records exceeds a predefined threshold stored as a data value in the machine-readable storage mechanism, then sampling a portion of the set of flat-data records and parsing the samples for storage into a training set, a test set, and a blind set by a machine-readable storage mechanism, and transformed to be accessible thereafter for use in the remaining steps, all with an apparatus for executing a sampling algorithm that limits the number of records that must be processed; 
 enhancing, enriching, and concentrating the sampled or unsampled data stored in the flat-data records with an apparatus for executing a data enrichment algorithm that removes data in particular data fields and the data fields themselves that are empirically predetermined to contribute little to building a predictive model, and that calculates and combines any data into new data fields that are predetermined to contribute most to building said predictive model, and that converts any text into a context classification with an apparatus for executing a context mining algorithm, and then transforming a plurality of intermediate results into a set of enriched-data records that are collectively stored in a machine-readable storage mechanism; 
 building a plurality of smart-agent predictive models for each entity identifiable in the samples of the enriched-data records with an apparatus for executing a smart-agent building algorithm, wherein such smart-agent predictive models have a class-prediction output and a confidence output; 
 selecting amongst the plurality of smart-agent predictive models obtained in the building and updating a corresponding particular smart-agent's real-time profile and long-term profile stored in a machine-readable storage mechanism with data from samples of the enriched-data records that correspond to a transaction activity of a particular entity with an apparatus for executing a smart-agent algorithm that compares a current transaction to previously memorialized transactions, and transforming and storing a result of that as smart-agent predictive model in a markup language document in a machine-readable storage mechanism; and 
 training at least one other predictive model technology for operation in parallel with the plurality of smart-agent predictive models, and that has its own independent class-prediction output and a confidence output; 
   thereafter:   receiving from a user-service consumer with a network input a series of informational electronic records that represent an encryption of business records, each business record having a predefined plurality of data fields, and structured and unstructured text, and then decrypting the business records one-by-one with an apparatus for executing a decoding algorithm, and then transforming a series of decrypted business record results into a set of non-transitory raw-data records that are collectively stored in a machine-readable storage mechanism as raw transaction data;   cleaning and improving the integrity of each raw transaction data record with an apparatus for executing a data integrity analysis algorithm that compares and corrects data values in each of their data fields, and that deduces the context of any included structured and unstructured text with an apparatus for executing a contextual dictionary algorithm, and that transforms each result into a non-transitory, cleaned, raw transaction data record that is stored in a machine-readable storage mechanism;   enhancing, enriching, and concentrating non-transitory, cleaned, raw transaction data record stored in the machine-readable storage mechanism with an apparatus for executing a data enrichment algorithm that removes data in particular data fields, and the data fields themselves, that are empirically predetermined to contribute little to assisting a predictive model, and that calculates and combines any data into new data fields that are predetermined to contribute most to assisting said predictive model, and that converts any text into a context classification with an apparatus for executing a context mining algorithm, and then transforming an intermediate result into an enriched business transaction data record that is stored in the machine-readable storage mechanism;   applying in parallel each enriched business transaction data record stored in the machine-readable storage mechanism to a selected one of the plurality of smart-agent predictive models and any and all other predictive model technologies such that each produce a prediction class output and a confidence output for the single enriched business transaction data record;   deciding which prediction class output to adopt as a final decision for output with an apparatus for executing an algorithm that compares the respective confidence outputs to threshold data values stored in the machine-readable storage mechanism according to an ordered set of rules in which at least one rule is provided for each predictive model; and   outputting with a network server an encrypted decision produced in the step of deciding to the user-service consumer.   
     
     
         10 . The method of  claim 9  wherein the step of training one other predictive model further comprises:
 building a data mining predictive model by applying the same data from the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a data mining algorithm, and transforming a result into a data-mining predictive model that is stored in a machine-readable storage mechanism. 
 
     
     
         11 . The method of  claim 9  wherein the step of training one other predictive model further comprises:
 building a neural network predictive model by applying the same data from the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a neural network algorithm, and transforming a result into a neural-network predictive model that is stored in a machine-readable storage mechanism. 
 
     
     
         12 . The method of  claim 9  wherein the step of training at least one other predictive model further comprises:
 building a case-based-reasoning predictive model by applying the same data from the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a cased-based reasoning algorithm, and transforming a result into a case-based-reasoning predictive model that is stored in a machine-readable storage mechanism. 
 
     
     
         13 . The method of  claim 9  wherein the step of training at least one other predictive model further comprises:
 building a clustering predictive model by applying the same data from the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a clustering algorithm, and transforming a result into a clustering predictive model that is stored in a machine-readable storage mechanism. 
 
     
     
         14 . The method of  claim 9  wherein the step of training at least one other predictive model further comprises:
 building a business rules predictive model by applying the same data from the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a business rules algorithm, and transforming a result into a business rules predictive model that is stored in a machine-readable storage mechanism. 
 
     
     
         15 . The method of  claim 9  wherein the step of training at least one other predictive model further comprises:
 building a data mining predictive model capable of scoring a record and assigning a confidence to a corresponding score by applying the same data from the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a data mining algorithm, and transforming a result into a data-mining predictive model that is stored in a machine-readable storage mechanism; 
 building a neural network predictive model capable of scoring a record and assigning a confidence to a corresponding score by applying the same data from the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a neural network algorithm, and transforming a result into a neural-network predictive model that is stored in a machine-readable storage mechanism; 
 building a case-based-reasoning predictive model capable of scoring a record and assigning a confidence to a corresponding score by applying the same data from the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a cased-based reasoning algorithm, and transforming a result into a case-based-reasoning predictive model that is stored in a machine-readable storage mechanism; 
 building a clustering predictive model capable of scoring a record and assigning a confidence to a corresponding score by applying the same data from the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a clustering algorithm, and transforming a result into a clustering predictive model that is stored in a machine-readable storage mechanism; and 
 building a business rules predictive model capable of scoring a record and assigning a confidence to a corresponding score by applying the same data from the enriched-data records corresponding to said transaction activity as an input to an apparatus for generating a business rules algorithm, and transforming a result into a business rules predictive model that is stored in a machine-readable storage mechanism.