User behavior models based on source domain
A method for tailoring content in a web page is provided. There is a relationship between the source domain a user comes from and the behavior pattern of a user on a website. To predict the behavior patterns of a particular user coming from a particular source domain, first a large dataset is compiled from user logs. Second, session analysis is performed on the dataset to identify sessions, session characteristics, page view categories, and source categories. Third, sessions are clustered and analyzed to determine how the behavior changes according, to a source category. Fourth, a mapping between source categories and behavior patterns is generated. When a user requests a page view from a source domain, if the source domain is mapped to a source category, then a tailored version of the page view is presented to the user based on the mapping between source categories and behavior patterns.
1. A method implemented on a machine having at least one processor, storage, and a communication platform connected to a network for content recommendation, the method comprising:
maintaining a past-behavior dataset that indicates how previous visitors to a target site have interacted at the target site, wherein the past-behavior dataset is organized based on establishing mappings between source domains from which the previous visitors visited the target site and a plurality of past behavior patterns, wherein each of the previous visitors engage with the site in any of the plurality of past behavior patterns;
in response to a user visiting the target site, performing the steps of:
determining a source domain from which the user is visiting the target site;
selecting, a particular past behavior pattern of the plurality of past behavior patterns based on the source domain being mapped to the particular past behavior pattern; and
tailoring a format of content to be provided to the user at the target site based on the selected past particular behavior pattern.
2. The method of claim 1 , wherein prior to the user visiting the target site, the method comprising:
obtaining one or more logs that indicate a set of page views of pages from the target site;
identifying a plurality of sessions that are reflected in the one or more logs; and
generating the mapping from the source domain to the particular past behavior pattern based on (a) a mapping from the source domain to a source category and (b) a mapping from the source category to the particular past behavior pattern.
3. The method of claim 2 , wherein the mapping from the source category to the particular past behavior pattern is based on the source category associated with a session-cluster, wherein the session-cluster comprises one or more sessions of the plurality of sessions, and the particular past behavior pattern associated with the session-cluster, wherein the source category is mapped to one or more of source domains.
4. The method of claim 3 , wherein the source category is associated with the session-cluster based on (a) a first contribution of the session-cluster in the source category and (b) a second contribution of the session-cluster in the particular past behavior pattern.
5. The method of claim 3 , further comprising generating the session-cluster from one or more sessions of the plurality of sessions based on canopy clustering.
6. The method of claim 3 , further comprising generating the session-cluster from one or more sessions of the plurality of sessions based on K-means clustering.
7. The method of claim 3 , further comprising determining an entropy for the session-cluster;
wherein the mapping from the source category to the particular past behavior pattern is based on the entropy for the session-cluster being greater than a particular threshold.
8. The method of claim 2 , wherein:
each of the set of page views includes one or more fields; and
the method comprises filtering the set of page views based on content of at least one of the one or more fields of each of the set of page views.
9. The method of claim 2 , wherein the mapping from the source domain to the source category is established manually.
10. The method of claim 2 , further comprising:
assigning one or more of the set of page views to a page view category based on layout of the set of page views;
wherein the particular past behavior pattern corresponds to the page view category.
11. The method of claim 1 , wherein tailoring the format of the content includes accentuating a portion of the content to be provided to the user.
12. A non-transitory machine-readable medium having information recorded thereon for content recommendation, wherein the information, when read by the machine, causes the machine to perform:
maintaining a past-behavior dataset that indicates how previous visitors to a target site have interacted at the target site, wherein the past-behavior dataset is organized based on establishing mappings between source domains from which the previous visitors visited the target site and a plurality of past behavior patterns, wherein each of the previous visitors engage with the site in any of the plurality of past behavior patterns;
in response to a user visiting the target site, performing the steps of:
determining a source domain from which the user is visiting the target site;
selecting, a particular past behavior pattern of the plurality of past behavior patterns based on the source domain being mapped to the particular past behavior pattern; and
tailoring a format of content to be provided to the user at the target site based on the selected past particular behavior pattern.
13. The medium of claim 12 , wherein the instructions cause the one or more computing devices to perform, prior to the user visiting the target site:
obtaining one or more logs that indicate a set of page views of pages from the target site;
identifying a plurality of sessions that are reflected in the one or more logs; and
generating the mapping from the source domain to the particular past behavior pattern based on (a) a mapping from the source domain to a source category and (b) a mapping from the source category to the particular past behavior pattern.
14. The medium of claim 13 , wherein:
the mapping from the source category to the particular past behavior pattern is based on:
the source category associated with a session-cluster, wherein the session-cluster comprises one or more sessions of the plurality of sessions, and
the particular past behavior pattern associated with the session-cluster, wherein
the source category is mapped to one or more of source domains.
15. The medium of claim 14 , wherein the source category is associated with the session-cluster based on (a) a first contribution of the session-cluster in the source category and (b) a second contribution of the session-cluster in the particular past behavior pattern.
16. The medium of claim 14 , wherein the instructions cause the machine to perform generating the session-cluster from one or more sessions of the plurality of sessions based on canopy clustering.
17. The medium of claim 14 , wherein the instructions cause the machine to perform generating the session-cluster from one or more sessions of the plurality of sessions based on K-means clustering.
18. The medium of claim 14 , wherein:
the instructions cause the machine to perform determining an entropy for the session-cluster; and
the mapping from the source category to the particular past behavior pattern is based on the entropy for the session-cluster being greater than a particular threshold.
19. The medium of claim 13 , wherein:
each of the set of page views includes one or more fields; and
the method comprises filtering the set of page views based on content of at least one of the one or more fields of each of the set of page views.
20. The medium of claim 13 , wherein the mapping from the source domain to the source category is established manually.
21. The medium of claim 13 , wherein the instructions cause the machine to perform:
assigning one or more of the set of page views to a page view category based on layout of the set of page views;
wherein the particular past behavior pattern corresponds to the page view category.