Bipartite Link Prediction by Intra-Class Connection Based Triadic Closure

Jungwoon Shin,Jaewoo Kang,Mogan Gim,Sunkyu Kim,Donghyeon Park

doi:10.1109/access.2020.3010223

Jungwoon Shin, Jaewoo Kang + Show 3 more

Open Access

https://doi.org/10.1109/access.2020.3010223

Copy DOI

Abstract

A variety of real-world systems can be formulated as bipartite link prediction problems where two different types of nodes exist and no links connect nodes of the same type. In link prediction, triadic closure is an important property that describes how new links are formed. However, triadic closure is difficult to apply to bipartite link prediction tasks because the triadic closure property, which states that new edges tend to form triangles, does not hold true in bipartite settings. In this paper, we introduce Intra-class Connection based Triadic Closure (ICTC) which is a method that can use triadic closure even when the nodes in the same set are unconnected. ICTC aggregates the link probabilities of many local triads, which are edges between triples of nodes, to predict the probability of a link existing between nodes. Specifically, the probability of an edge in a triangle is calculated by multiplying the probabilities of two other edges. The experimental results on eight real-world datasets show that our method outperforms state-of-the-art methods in most cases.

Highlights

In various fields, many types of complex data representing relationships or interactions among entities can be represented as graphs [1], [2]
The baseline studies used for our experiments employ metrics such as precision and predictive power, these metrics suffer from a problem in specifying the number of top K instances for ranking the edges with the highest scores
If there are m times that the value measured from test set of edges is bigger than the value measured from test set of false edges and m times that the two values are equal, Area Under the Receiver Operating Characteristics (AUC-ROC) can be computed as follows: AUC-ROC = (m + 0.5m )/n

Summary

Introduction

Many types of complex data representing relationships or interactions among entities can be represented as graphs [1], [2]. There are numerous graph related machine learning tasks such as link prediction, node classification, or node clustering. Many researches have been conducted to solve the problem of link prediction in monopartite networks, which can be divided into two groups: generative-model and model-free. The generative-model is further separated into two groups: mechanistic model and probabilistic model. The mechanistic model includes Jaccard’s Index [3], [4], which is the earliest link prediction method, common neighbors [5], Adamic-Adar [6], resource allocation [7], and CRA [8]. The probabilistic model includes stochastic relational model [9], stochastic block model [10], and hierarchical structure mode [11]. The model-free models include structural perturbation method (SPM) [12]

Results

Discussion

Conclusion