Graph-based features for supervised link prediction

Abstract

The growing ubiquity of social networks has spurred research in link prediction, which aims to predict new connections based on existing ones in the network. The 2011 IJCNN Social Network challenge asked participants to separate real edges from fake in a set of 8960 edges sampled from an anonymized, directed graph depicting a subset of relationships on Flickr. Our method incorporates 94 distinct graph features, used as input for classification with Random Forests. We present a three-pronged approach to the link prediction task, along with several novel variations on established similarity metrics. We discuss the challenges of processing a graph with more than a million nodes. We found that the best classification results were achieved through the combination of a large number of features that model different aspects of the graph structure. Our method achieved an area under the receiver-operator characteristic (ROC) curve of 0.9695, the 2nd best overall score in the competition and the best score which did not de-anonymize the dataset.

Link Prediction in Social Networks Using Computationally Efficient Topological Features

“…It is expected that the larger the size of the common neighborhood, the higher the chances that both vertices will be connected. The common-friends feature was widely used in the past link prediction on several datasets, and found to be very helpful [3], [6], [8], [9], [10]. Total-Friends: For two vertices u, v, we can define the number of distinct friends that u and v have together, namely: Let be u, v ∈ V we define the total-friends of u, v to be the number of distinct neighbors u, v has:…”

“…Jaccard's coefficient: Jaccard's-coefficient is a well-known feature for link prediction [3], [6], [8], [9], [10]. The Jaccard coefficient, which measures the similarity between sample sets, is defined as the size of the intersection divided by the size of the union of the sample sets.…”

“…al. [10] extracted 94 distinct graph features. Using a Random Forests classifier, they achieved impressive results in predicting links on Flickr datasets.…”

The predictive value of young and old links in a social network

“…The link prediction problem can also be treated as a supervised classification problem in which the labels represent presence or absence of links and the features can be both topological (such as the shortest distance between a pair of nodes or the clustering index) and non-topological (such as the intrinsic properties of the nodes) [1]. Recently, Cukierski showed that using a large number of network structure based features for link prediction is promising [7]. Surveys of the link prediction problem can be found in [10,17].…”

“…While it is shown that performance of link prediction can be improved by introducing a large number of features [7], most of these features have nothing to do with links' ages. Introducing many features may improve precision but will make it difficult to isolate the effect of link age.…”