Methods and apparatus for managing models for classification of online users

A system for training classification models, the system comprising:

• an impression handler implemented by processor circuitry to receive an impression from a user computing device and to determine a segment associated with a received impression indicative of access to media on a webpage;
• a trainer implemented by the processor circuitry to:
• determine that the received impression includes an indication that the received impression is a converted impression that is associated with a conversion, and
• determine a current ratio of converted to non-converted impressions associated with a classification model associated with the segment;
• determine a difference between the current ratio and a target ratio;
• a datastore to store an indication of the received impression in a training list of the classification model associated with the segment when:
• (a) the difference does not meet a threshold or (b) counting the received impression in the current ratio would reduce the difference by including an additional converted impression in training data for the classification model;
• and
• a scorer implemented by the processor circuitry to:
• update the classification model using impressions identified in the training list to generate an updated classification model;
• receive a subsequent media access associated with an impression;
• generate, using the updated classification model, a score indicating whether the subsequent media access will result in a conversion;
• add the impression associated with the subsequent media access to the training list based on the score indicating that the impression is associated with a conversion;
• and
• update the classification model using the impression added to the training list to generate an updated classification model.

A non-transitory computer readable medium comprising instructions that, when executed, cause a machine to at least:

• determine a segment associated with a received impression indicative of access to media on a webpage;
• determine that the received impression includes an indication that the received impression is a converted impression that is associated with a conversion;
• determine a current ratio of converted to non-converted impressions associated with a classification model associated with the segment;
• determine a difference between the current ratio and a target ratio;
• store an indication of the received impression in a training list of the classification model associated with the segment when:
• (a) the difference does not meet a threshold or (b) counting the received impression in the current ratio would reduce the difference by including an additional converted impression in training data for the classification model;
• update the classification model using impressions identified in the training list to generate an updated classification model;
• receive a subsequent media access associated with an impression;
• generate, using the updated classification model, a score indicating whether the subsequent media access will result in a conversion;
• add the impression associated with the subsequent media access to the training list based on the score indicating that the impression is associated with a conversion;
• and
• update the classification model using the impression added to the training list to generate an updated classification model.