Group relationship type identification method and apparatus

Abstract

The invention discloses a group relationship type identification method and apparatus, and belongs to the field of data mining analysis. The group relationship type identification method comprises thesteps of receiving a group relationship type identification request; obtaining group information corresponding to the request; according to predefined target data fields, extracting target data corresponding to members in a group from the group information, thereby forming personal data sets; inputting the personal data sets to a group classification deep neural network model, wherein the group classification deep neural network model is obtained by training according to the predefined target data fields; and according to an output result of the group classification deep neural network model,determining a relationship type of the group. According to the technical scheme, the relationship type of the group can be identified through simple preprocessing operation and the group classification deep neural network model in combination with the data mining analysis technology, so that the manpower cost required for obtaining earlier stage features is reduced, the deployment and building complexity of the model is low, and the universality of the model is high.

Description

technical field [0001] The invention relates to the field of data mining and analysis, in particular to a group relationship type identification method and device. Background technique [0002] There are various social circles in the social network, and each social circle has certain related user members, such as family members, company colleagues, school classmates, and so on. In the context of the big data era, identifying the relationship types of social circles is a very important issue, and the identification results have a wide range of practical applications, such as big data analysis, advertising and so on. [0003] Existing technical solutions either use manual identification and classification, or use traditional machine learning classification models, which require a lot of feature engineering work at the community level. The required features include members of social circles, age and gender distribution of members, geographical distribution, etc. . Improvement...

AI Technical Summary

Problems solved by technology

[0004] 1> The completion of feature engineering requires a lot of manpower input, and the development cycle is long;

[0005] 2> Different circle classification scenarios require different feature engineering work, which has poor versatility;

[0006] 3> Due to the large amount of feature processing involved, the complexity of model deployment is high;

[0007] 4> The characteristics of the circle come from the statistics of personal characteristics, and a lot of information that is helpful to improve the accuracy rate is lost

Images