Method and device for social platform-based data mining

Patent metadata

Field	Value
Publication number	US-2017322981-A1
Application number	US-201515525870-A
Country	US
Kind code	A1
Filing date	Jul 10, 2015
Priority date	Nov 10, 2014
Publication date	Nov 9, 2017
Grant date	—

Abstract

Provided is a method and device for social platform-based data mining. The method includes: acquiring one or more interest label dictionaries of one or more registered users on an information client and one or more first objects having followed relationship with the one or more registered users on the information client in a social platform; determining one or more first followed sets corresponding to the one or more registered users; constructing an interest model; acquiring one or more second objects having followed relationship with one or more newly registered users on the information and reading relationship information between the one or more newly registered users and the one or more second objects; determining a second followed set corresponding to the one or more newly registered users; and matching the second followed set with the interest model to determine one or more recommended interest labels of the one or more newly registered users.

First claim

1 . A method for social platform-based data mining, comprising:
acquiring one or more interest label dictionaries of one or more registered users on an information client; acquiring one or more first objects having followed relationship with the one or more registered users on the information client in a social platform and reading relationship information between the one or more registered users and the one or more first objects; according to the one or more first objects having the followed relationship with the one or more registered users, determining one or more first followed sets corresponding to the one or more registered users; according to the one or more interest label dictionaries of the one or more registered users and the one or more first followed sets, constructing an interest model, wherein the interest model is used to characterize a corresponding relationship between one or more registered users having the same first followed set and an interest label; acquiring one or more second objects having followed relationship with one or more newly registered users on the information client in the social platform, and reading relationship information between the one or more newly registered users and the one or more second objects; according to the one or more second objects having the followed relationship with the one or more newly registered user, determining a second followed set corresponding to the one or more newly registered users; and matching the second followed set with the interest model to determine one or more recommended interest labels of the one or more newly registered users according to the interest model.

2 . The method as claimed in claim 1 , wherein before acquiring the one or more interest label dictionaries of the one or more registered users on the information client, the method comprises:
acquiring recommended information; extracting one or more interest labels of the recommended information from content of the recommended information; acquiring historical behavior data of the one or more registered users, wherein the historical behavior data is used to record operational behavior of the one or more registered users for the recommended information; determining one or more label weight values of the one or more interest labels according to the historical behavior data; and determining the one or more interest label dictionaries corresponding to the one or more registered users according to the one or more label weight values.

3 . The method as claimed in claim 2 , wherein constructing the interest model according to the one or more interest label dictionaries of the one or more registered users and the first followed set comprises:
performing filtering in the first followed set to acquire a third followed set corresponding to the one or more registered users, wherein a method for performing filtering in the first followed set comprises at least one of a data filtering method, an index filtering method, a condition filtering method, and an information filtering method; matching the one or more registered users based on the third followed set to generate a registered user set, wherein the registered user set comprises one or more registered users having the same third followed set; and generating a user set label dictionary corresponding to the registered user set according to the one or more interest label dictionaries of the one or more registered users comprised in the registered user set.

4 . The method as claimed in claim 3 , wherein generating the user set label dictionary corresponding to the registered user set according to the one or more interest label dictionaries of the one or more registered users comprised in the registered user set comprises:
acquiring a first user amount of the one or more registered users on the information client and a second user amount of the registered user set; calculating a weight distribution average value of each of the interest labels according to the one or more label weight values and a first user amount; calculating a set weight average value of each of the interest labels in a user set interest label dictionary according to the one or more label weight values and a second user amount of the one or more registered users in the registered user set; calculating a registered user set weight value of the one or more interest labels in the user set interest label dictionary according to the weight distribution average value and the set weight average value; successively comparing the registered user set weight value of the one or more interest labels in the user set interest label dictionary with a preset noise threshold; retaining the interest label corresponding to the registered user set weight value in the user set label dictionary when the registered user set weight value of the one or more interest labels in the user set interest label dictionary is greater than the preset noise threshold; and deleting the interest label corresponding to the registered user set weight value from the user set label dictionary when the registered user set weight value of the one or more interest labels in the user set interest label dictionary is less than or equal to the preset noise threshold.

5 . The method as claimed in claim 4 , wherein matching the second followed set with the interest model to determine one or more recommended interest labels of the one or more newly registered users according to the interest model comprises:
performing filtering in the second followed set to acquire a fourth followed set corresponding to the one or more newly registered users, wherein the filtering method comprises at least one of a data filtering method, an index filtering method, a condition filtering method, and an information filtering method; matching the fourth followed set with the third followed set to determine the registered user set corresponding to the one or more newly registered users; and determining the one or more recommended interest labels of the one or more newly registered users according to the user set label dictionary of the registered user set corresponding to the one or more newly registered users.

6 . The method as claimed in claim 1 , wherein after matching the second followed set with the interest model to determine one or more recommended interest labels of the one or more newly registered users according to the interest model, the method further comprises:
pushing recommended information for the one or more newly registered users according to the one or more recommendation interest labels.

7 . The method as claimed in claim 4 , wherein the registered user set weight value V′[i] of the one or more interest labels is determined by the following formula:
V′[i]=V[i]/V base [i];
where V′[i] denotes the registered user set weight value of an interest label i, V[i] denotes the set weight average value of the interest label i, V base [i] denotes the weight distribution average value of the interest label i.

8 . A device for social platform-based data mining, comprising:
a first acquiring component arranged to acquire one or more interest label dictionaries of one or more registered users on an information client; a second acquiring component arranged to acquire one or more first objects having followed relationship with the one or more registered users on the information client in a social platform and read the relationship information between the one or more registered users and the one or more first objects; a first determining component arranged to, according to the one or more first objects having the followed relationship with the one or more registered users, determine one or more first followed sets corresponding to the one or more

Assignees

• Beijing Bytedance Network Tech Co Ltd

Inventors

• Zhang Yiming
• Chen Tao
• Cao Huanhuan
• Luo Lixin

Classifications

• G06F16/24578

  using ranking · CPC title

• G06F16/9535 Primary

  Search customisation based on user profiles and personalisation · CPC title

• G06F16/00

  Information retrieval; Database structures therefor; File system structures therefor · CPC title

• G06F16/2465 Primary

  Query processing support for facilitating data mining operations in structured databases · CPC title

• G06F16/24575

  using context · CPC title