Dynamic model data facility and automated operational model building and usage

A non-transitory computer-readable medium comprising processor-executable instructions that, when executed by a processor, cause the processor to perform a method that comprises:

• building a model data file, the model data file including normalized data values from two or more data streams that have been joined and aggregated across one or more periods of time;
• scoring, in real-time, data received from the two or more data streams;
• as the data is being scored in real-time, determining one or more instances of matching content;
• storing results and scoring associated with the model data file in a clustered cache;
• and
• dynamically updating, in real-time, the model data file based on an analysis of modeled data and the scoring provided by the clustered cache provided by the model data file being used in at least one expert system, wherein the dynamically updating is in real-time by periodically re-calibrating coefficients for the model data file at predetermined intervals by using a moving data horizon then re-calibrating coefficients for an incumbent model and re-specifying the model data file based on one or more p-values that are generated using the model data file, wherein re-specifying the model data file comprises generating a new model data file to replace the model data file and wherein the new model data file is generated by re-running feature creation and feature selection on a most recent data set, wherein the normalized data values are calculated based on raw data obtained from at least one of the two or more data streams and wherein the one or more p-values are selected from a constellation of p-values that include p-values for calibration and p-values for differing cut-points for True Positive Rate (TPR) and False Positive Rate (FPR).

A system, comprising:

• a server comprising memory and a processor coupled thereto, the memory including instructions that are executable by the processor, the instructions including:
• instructions that enable the processor to receive two or more streams of data from different data streams, wherein the two or more streams of data are received in different formats from one another;
• instructions that enable the processor to normalize data values from the two or more streams of data in real-time;
• instructions that enable the processor to join and aggregate the normalized data values across one or more historical periods of time;
• instructions that enable the processor to build a model data file with the normalized data values after the normalized data values have been joined and aggregated across the one or more historical periods of time;
• instructions that score, in real-time, data received from the two or more data streams;
• instructions that determine one or more instance of matching content as the data is being scored in real-time;
• instructions that store results and scoring associated with the model data file in a clustered cache;
• and
• instructions that enable the processor to dynamically updating, in real-time, the model data file based on an analysis of modeled data and the scoring provided by the clustered cache provided by the model data file being used in at least one expert system, wherein the dynamically updating is in real-time by periodically re-calibrating coefficients for the model data file at predetermined intervals by using a moving data horizon then re-calibrating coefficients for an incumbent model and re-specifying the model data file based on one or more p-values that are generated using the model data file, wherein re-specifying the model data file comprises generating a new model data file to replace the model data file and wherein the new model data file is generated by re-running feature creation and feature selection on a most recent data set, wherein the normalized data values are calculated based on raw data obtained from at least one of the two or more data streams and wherein the one or more p-values are selected from a constellation of p-values that include p-values for calibration and p-values for differing cut-points for True Positive Rate (TPR) and False Positive Rate (FPR).