|
Predictive behavioral targeting uses a linking of surveys and measurement data to open up the entire spectrum for behavioral targeting. The Predictive Behavioral Targeting system learns from user behavior combined with survey (concerning socio-demographics, product interests and lifestyle) or other third party data in real time. Machine-learning algorithms are put to work in order to provide ad servers with precise profile information for the whole inventory. The technology used is the same as in research on artificial intelligence and robotics. == Methodology == The methodology is based on measurement data for online usage, enriched with information gathered through a survey of sampled users. In a nutshell, the methodology encompasses three steps: # Cookies are saved on the computers of all users of a portal or marketing network. These cookies indicate how often the users have visited certain websites (measurement). # A random sample of users is polled on their demographics, interests and lifestyle (surveys). # This information is overlaid - online and in real time - onto the entirety of the user group (projection). This process provides a complete targeting profile containing both product interests on the basis of visited online content as well as indications of demographics, interests and lifestyle. Survey data is projected onto the entirety of users by forming "statistical twins": Users without survey data "inherit" the missing survey data from those surveyed users whose measured surfing behavior most closely resembles their own. 抄文引用元・出典: フリー百科事典『 ウィキペディア（Wikipedia）』 ■ウィキペディアで「Predictive behavioral targeting」の詳細全文を読む スポンサード リンク
|