Automated predictive data mining model selection

One or more computer-readable media comprising computer-executable instructions causing a computer to perform a computer-implemented method comprising:

• receiving a set of candidate predictive data mining model templates;
• repeatedly transforming the set of candidate predictive data mining model templates, wherein the transforming comprises building predictive data mining models based on templates in the set and historical data, scoring the predictive data mining models based on test data and an objective function, and choosing, by software, based on the scoring, additional candidate predictive data mining model templates for inclusion in the set based on scoring results;
• building and scoring additional predictive data mining models based on the additional candidate predictive data mining model templates;
• and
• storing an indication of a candidate predictive data mining model template for which a built and scored predictive data mining model best meets the objective function.

A computer-implemented method comprising:

• inputting a first dataset comprising historical data;
• receiving a set of predictive data mining model templates;
• building a plurality of predictive data mining models based on the first dataset and the set of predictive data mining model templates,
• inputting a second dataset comprising test data;
• scoring the plurality of predictive data mining models via applying the second dataset to the plurality of predictive data mining models;
• repeatedly transforming the set of predictive data mining model templates based on the scoring, the transforming comprising building additional predictive data mining models in the plurality of predictive data mining models and scoring the additional predictive data mining models;
• and
• storing an indication of a predictive data mining model best meeting an objective function out of the plurality of predictive data mining models based on the scoring of the plurality of predictive data mining models.

One or more computer-readable media having encoded thereon a predictive data mining modeling system comprising:

• an inputted historical dataset,
• an inputted test dataset;
• an inputted set of predictive data mining model templates;
• an inputted at least one objective function;
• a predictive data mining model building module operable to build a plurality of predictive data mining models based at least in part on the historical dataset and a first predictive data mining model template selected from the set of predictive data mining model templates;
• a predictive data mining model scoring module operable to score the plurality of predictive data mining models based on an application of the test dataset to the plurality of predictive data mining models;
• a predictive data mining model template transformation module operable to repeatedly transform the plurality of predictive data mining model templates;
• and
• a predictive data mining model selection module operable to select a predictive data mining model from the plurality of predictive data mining models based on results provided by the predictive data mining model scoring module.

One or more computer-readable media having encoded thereon a digital chromosome representing characteristics of a predictive data mining model, the digital chromosome comprising:

• a first field indicating a classifier type of the predictive data mining model;
• a second field indicating a learning scheme of the predictive data mining model;
• and
• a third field indicating control options for the predictive data mining model;
• wherein the digital chromosome is sufficient to cause creation of a predictive data mining model having the indicated characteristics when submitted with historical data to a predictive data mining model builder.