Cross-social media account matching approach that integrates social relationships and naming features

Abstract

The invention discloses a cross-social media account matching method which integrates social relations and naming characteristics. The method comprises the following steps: S1, finding out the users corresponding to the accounts of other platforms in the account information as a set of seed users; S2, extracting a pair of account information from the seed user set, respectively extracting the account number of a friend in the corresponding platform, and making a Cartesian product as a candidate account pair; S3, carrying out pretreatment to obtain features to form feature vector; S4, inputtingthe feature vector into the classifier for judging, and adding the account judged to belong to the same person entity into the seed node set; S5, performing the operations of steps S2 to S4 on all the seed nodes until no new account is added in the seed user set. The method of the invention reduces the computational complexity, and the calculation of the extracted features of the users which arefinally judged to be unrelated and carried out classification judgment is called invalid calculation, which greatly reduces the proportion of invalid calculation in the whole calculation process and improves the efficiency.

Description

technical field [0001] The invention relates to a cross-social media account matching method that integrates social relations and naming features. Background technique [0002] For the account matching problem across social media, it is usually modeled as a classification problem: given two users a and b belong to social media platforms A and B respectively, the purpose of the research is to obtain a classification discriminant function, when the input user After the information of a and b, if the two users belong to the same person entity, output 1, otherwise output 0, as shown in the following publicity: [0003] [0004] The method of character recognition across social media is of great significance to the research of data mining, and can be used as the premise of various researches. [0005] The main idea of ​​the existing cross-social media user matching method is to convert user matching into a binary classification problem, combine various features such as docume...

AI Technical Summary

Problems solved by technology

[0006] First, the generalization ability of the classification model is poor

Existing research mainly focuses on feature construction and model selection. For the construction of training sets and test sets, users who actively provide other social platform accounts in their accounts are selected as positive examples, and on this basis, the known matching relationships in positive examples are used as positive examples. Randomly scramble to obtain negative examples. The disadvantage of this method is that the negative example data is more random, and the number of samples at the decision boundary in the training set is small, which leads to the inaccuracy of the constructed classifier.

[0007] Second, it cannot be applied to practical applications

The usage scenario of cross-social media user matching is usually a set of known different social media accounts, and it is necessary to match the accounts corresponding to each other in the two sets. Compared with each other in turn, there is a problem of high computational complexity, and this method cannot be extended to the actual mass social media account matching

Images