Methods and apparatus to profile geographic areas of interest

A tangible computer readable storage medium, comprising instructions that, when executed, cause a machine to at least:

• identify a first geographic area for which a model database does not include a machine learning model, the first geographic area including a first data element;
• identify a first set of trained machine learning models corresponding to a second geographic area, the second geographic area including a second data element that matches the first data element of the first geographic area, the second geographic area including a first affinity corresponding to the first geographic area, the first set of trained machine learning models to predict consumer information for the second geographic area;
• identify a second set of trained machine learning models corresponding to a third geographic area, the third geographic area including a third data element that matches the first data element of the first geographic area, the third geographic area including a second affinity corresponding to the first geographic area, the second set of trained machine learning models to predict consumer information for the third geographic area;
• when a first one of the first set of trained machine learning models conflicts with the second set of trained machine learning models, mix (a) non-conflicting ones of the first set of trained machine learning models and (b) the second set of trained machine learning models to generate a composite set of machine learning models, the composite set of machine learning models excluding the first one of the first set of trained machine learning models that conflicted with the second set of trained machine learning models based on a determination that the first affinity is less than the second affinity;
• and
• in response to determining that the composite set of machine learning models represents the first geographic area, reduce computational complexity of profiling the first geographic area by using the composite set of machine learning models to represent the first geographic area in the model database with a third machine learning model trained by the composite set of machine learning models, the composite set of machine learning models predicting consumer information for the first geographic area.

A hardware-implemented apparatus, comprising:

• a geographic area manager to determine a first data element of a first geographic area;
• a model searcher to determine whether a model database includes a first set of trained machine learning models corresponding to a second geographic area and a second set of trained machine learning models corresponding to a third geographic area, the first set of trained machine learning models to predict consumer information for the second geographic area, the second set of trained machine learning models to predict consumer information for the third geographic area;
• a geographic area matcher to determine (a) whether the second geographic area includes a second data element that matches the first data element of the first geographic area and (b) whether the third geographic area includes a third data element that matches the first data element of the first geographic area, the second geographic area including a first affinity corresponding to the first geographic area, the third geographic area including a second affinity corresponding to the first geographic area;
• a trained model extractor to extract, from the model database, the first set of trained machine learning models corresponding to the second geographic area and the second set of trained machine learning models corresponding to the third geographic area;
• and
• a model mixer to:
• when a first one of the first set of trained machine learning models conflicts with the second set of trained machine learning models, blend (a) non-conflicting ones of the first set of trained machine learning models and (b) the second set of trained machine learning models together into a blended set of machine learning models to represent the first geographic area, the blended set of machine learning models excluding the first one of the first set of trained machine learning models that conflicted with the second set of trained machine learning models based on a determination that the first affinity is less than the second affinity;
• and
• in response to determining that the blended set of machine learning models represents the first geographic area, reduce computational complexity of profiling the first geographic area with a third machine learning model by inserting into a first dataset for the first geographic area, data from the blended set of machine learning models such that an amount of information in the first dataset satisfies a threshold amount of information, the third machine learning model predicting consumer information for the first geographic area.