A kind of feature extraction based on user behavior, personalized recommendation method and system

A kind of feature extraction based on user behavior, personalized recommendation method and system

  • CN 104,239,324 B
  • Filed: 06/17/2013
  • Issued: 09/17/2019
  • Est. Priority Date: 06/17/2013
  • Status: Active Grant
First Claim
Patent Images

1. a kind of feature extracting method based on user behavior characterized by comprisingThe primitive behavior information for collecting user, the primitive behavior information according to the user generate user behavior data point;

  • It is describedUser behavior data point is the vector point of hyperspace, including user identifier, dimension mark and corresponding dimension score value three toSecondary element;

    The user behavior data point is clustered using dimension score value according to dimension mark, it is empty to form multiple clustersBetween;

    Binaryzation is carried out for the dimension score value of user behavior data point in the multiple Cluster space respectively;

    Result according to binaryzation extracts feature dimensions of the one or more dimensions mark for meeting preset requirement as Cluster spaceDegree;

    Wherein, described that the user behavior data point is clustered using dimension score value according to dimension mark, it is formed multiple poly-The step of space-like, is applied to distributed computing, comprising;

    The user behavior data point is distributed into K cluster, and according to dimension mark calculate separately in K cluster it is corresponding initiallyK central point form K initial Cluster space and according to the K central point and corresponding user behavior data point;

    Wherein, the K is positive integer;

    Each user behavior data point is calculated separately at a distance from K central point according to dimension mark;

    According to the user behavior data point at a distance from K central point, the user behavior data point ownership is redefinedCluster space;

    Judge whether current K Cluster space meets preset condition;

    If so, obtaining K final Cluster space;

    If it is not, then being returned after the sub-step for executing the new central point for calculating separately current K Cluster space according to dimension markIt returns described identify according to dimension and calculates separately sub-step of each user behavior data point at a distance from K central point.

View all claims
    ×
    ×

    Thank you for your feedback

    ×
    ×