A method for constructing a fair link prediction evaluation system

Abstract

The invention discloses a method for constructing an evaluation system for fairly and effectively evaluating different link prediction methods. A general method for automatic collection, selection andgeneration of test data sets is proposed, which includes two algorithms: hierarchical clustering-based binary clustering tree construction algorithm and deepest leaf node-first data set selection algorithm. The invention presents a new method to measure the difficulty coefficients of data sets based on network topology indices, in which the optimal subset regression algorithm is used to measure the difficulty coefficients scientifically; provides a set of easy-to-use API design standards and implementation for accessing interfaces and manipulating datasets so that the new link prediction method can be easily evaluated. Compared with other evaluation methods, the method of the invention can fairly compare the performance of different link prediction methods, and can effectively evaluate the universality of the link prediction methods.

Description

Technical field [0001] The present invention relates to the field of evaluation system construction, in particular to a method for constructing a fair link prediction evaluation system. Background technique [0002] As one of the hot issues in the field of data mining, social network link prediction has important application prospects in the fields of friend recommendation in online social networks and product recommendation in e-commerce. Applying the technology of link prediction can help companies tap potential social relationships, increase user stickiness, and enhance the business value that company users bring to the company. [0003] So far, many social network link prediction technologies and related works based on different perspectives have been proposed. Although these methods and works all claim that the link prediction technology proposed by themselves has good performance, because there is no fair and effective unified evaluation platform, it is impossible to objecti...

AI Technical Summary

Problems solved by technology

Since the experiment shows that the selection of the evaluation data set has a great influence on the experimental results of the link prediction method, the lack of this systematic and objective process will lead to the unfairness of using the existing technology to evaluate different link prediction methods

Moreover, the existing technology lacks research on the generality of link prediction methods, that is, there is a lack of analysis of whether a specific link prediction method will show performance differences on different types of network data sets

Images