Impression-trend techniques for providing a display advertising supply forecast

A system to link impressions to trends, comprising:

• an aggregation module comprising electrical circuitry, the aggregation module being configured to aggregate impressions in each of a plurality of first subsets of impressions that corresponds to a plurality of respective time periods to provide a first periodically aggregated time series, each of the impressions in the first subsets having a first combination of attribute values that is related to a designated combination of attribute values, the aggregation module further configured to aggregate impressions in each of a plurality of second subsets of impressions that corresponds to the plurality of respective time periods to provide a second periodically aggregated time series, each of the impressions in the second subsets having a second combination of attribute values that is related to the designated combination of attribute values;
• a forecast module configured to estimate a first number of impressions in each of the first subsets that correspond to a first portion of the time periods using a first trend that is based on the first subsets that correspond to a second portion of the time periods to provide respective first forecasts, the forecast module further configured to estimate a second number of impressions in each of the second subsets that correspond to the first portion of the time periods using a second trend that is based on the second subsets that correspond to the second portion of the time periods to provide respective second forecasts;
• a comparison module configured to compare the first numbers and the first forecasts to determine a first cumulative error value, the comparison module further configured to compare the second numbers and the second forecasts to determine a second cumulative error value;
• and
• a linking module configured to link each impression that occurs in a specified period of time and that is included among the first subsets and among the second subsets to a third trend that is based on the first subsets that correspond to the first portion of the time periods and the first subsets that correspond to the second portion of the time periods if the first cumulative error value is less than the second cumulative error value, the linking module further configured to link each impression that occurs in the specified period of time and that is included among the first subsets and among the second subsets to a fourth trend that is based on the second subsets that correspond to the first portion of the time periods and the second subsets that correspond to the second portion of the time periods if the first cumulative error value is greater than the second cumulative error value, the linking module further configured to link each impression that occurs in the specified period of time and that is included among the first subsets and not among the second subsets to the third trend, the linking module further configured to link each impression that occurs in the specified period of time and that is included among the second subsets and not among the first subsets to the fourth trend.

A computer program product comprising a computer-readable storage medium having computer program logic recorded thereon for enabling a processor-based system to link impressions to trends, the computer program product comprising:

• a first program logic module for enabling the processor-based system to aggregate impressions in each of a plurality of first subsets of impressions that corresponds to a plurality of respective time periods to provide a first periodically aggregated time series,
• the first program logic module further for enabling the processor-based system to aggregate impressions in each of a plurality of second subsets of impressions that corresponds to the plurality of respective time periods to provide a second periodically aggregated time series,
• each of the impressions in the first subsets having a first combination of attribute values that is related to a designated combination of attribute values,
• each of the impressions in the second subsets having a second combination of attribute values that is related to the designated combination of attribute values;
• a second program logic module for enabling the processor-based system to estimate a first number of impressions in each of the first subsets that correspond to a first portion of the time periods using a first trend that is based on the first subsets that correspond to a second portion of the time periods to provide respective first forecasts,
• the second program logic module further for enabling the processor-based system to estimate a second number of impressions in each of the second subsets that correspond to the first portion of the time periods using a second trend that is based on the second subsets that correspond to the second portion of the time periods to provide respective second forecasts;
• a third program logic module for enabling the processor-based system to compare the first numbers and the first forecasts to determine a first cumulative error value,
• the third program logic module further for enabling the processor-based system to compare the second numbers and the second forecasts to determine a second cumulative error value;
• and
• a fourth program logic module for enabling the processor-based system to, for each impression that occurs in a specified period of time, link the impression to a third trend that is based on the first subsets that correspond to the first portion of the time periods and the first subsets that correspond to the second portion of the time periods if the impression is included among the first subsets and among the second subsets and the first cumulative error value is less than the second cumulative error value,
• the fourth program logic module further for enabling the processor-based system to, for each impression that occurs in the specified period of time, link the impression to a fourth trend that is based on the second subsets that correspond to the first portion of the time periods and the second subsets that correspond to the second portion of the time periods if the impression is included among the first subsets and among the second subsets and the first cumulative error value is greater than the second cumulative error value,
• the fourth program logic module further for enabling the processor-based system to, for each impression that occurs in the specified period of time, link the impression to the third trend if the impression is included among the first subsets and not among the second subsets,
• the fourth program logic module further for enabling the processor-based system to, for each impression that occurs in the specified period of time, link the impression to the fourth trend if the impression is included among the second subsets and not among the first subsets.