Automated processing of multiple prediction generation including model tuning

A method, comprising:

• receiving, by one or more processors, a data set, the dataset comprising a plurality of keys and a plurality of key-value relationships, wherein a key of the dataset corresponds to a grouping of columns of the dataset;
• determining a plurality of models to build based at least in part on the dataset, wherein determining the plurality of models to build comprises using a dataset format information to identify the plurality of models, wherein the determining of the plurality of models comprises:
• analyzing the dataset;
• extracting a subset of the plurality of keys, wherein the subset of the plurality of keys includes a set of all unique combinations of keys of the dataset;
• and
• determining the plurality of models using the subset of the plurality of keys, wherein each unique combination of keys of the dataset corresponds to a model;
• analyzing the dataset for which the corresponding model is built based on a unique dimension along which the dataset is analyzed for the each unique combination of keys for the dataset;
• building the plurality of models, comprising:
• providing the dataset to train the plurality of models in parallel, one model per each grouping of dataset;
• and
• providing a unified API to interface with each model of the plurality of models;
• optimizing at least one of the plurality of models;
• receiving, from a user, a request for a prediction using the unified API, the request for the prediction comprising one or more prediction parameters;
• determining a prediction model based at least in part on the one or more prediction parameters, the prediction model being selected from among the plurality of models, wherein the determining of the prediction model comprises:
• using the prediction model to generate the prediction;
• and
• providing the prediction to the user.

A computer program product embodied in a non-transitory computer readable medium and comprising computer instructions for:

• receiving, by one or more processors, a data set, the dataset comprising a plurality of keys and a plurality of key-value relationships, wherein a key of the dataset corresponds to a grouping of columns of the dataset;
• determining a plurality of models to build based at least in part on the dataset, wherein determining the plurality of models to build comprises using a dataset format information to identify the plurality of models, wherein the determining of the plurality of models comprises:
• analyzing the dataset;
• extracting a subset of the plurality of keys, wherein the subset of the plurality of keys includes a set of all unique combinations of keys of the dataset;
• and
• determining the plurality of models using the subset of the plurality of keys, wherein each unique combinations of keys of the dataset corresponds to a model;
• analyzing the dataset for which the corresponding model is built based on a unique dimension along which the dataset is analyzed for the each unique combination of keys for the dataset;
• building the plurality of models, comprising:
• providing the dataset to train the plurality of models in parallel, one model per each grouping of dataset;
• and
• providing a unified API to interface with each model of the plurality of models;
• optimizing at least one of the plurality of models,
• receiving, from a user, a request for a prediction using the unified API, the request for the prediction comprising one or more prediction parameters;
• determining a prediction model based at least in part on the one or more prediction parameters, the prediction model being selected from among the plurality of models, wherein the determining of the prediction model comprises:
• using the prediction model to generate the prediction;
• and
• providing the prediction to the user.