China
Africa
Explore chapters and articles related to this topic
Intelligent Social Networking in CPS
Published in G.R. Karpagam, B. Vinoth Kumar, J. Uma Maheswari, Xiao-Zhi Gao, Smart Cyber Physical Systems, 2020
S. Hemkiran, G. Sudha Sadasivam
The interactions among the multitude of constituent nodes in a social network leads to the formation of communities with shared interests. In order to form communities or groups, it is essential to identify entities with similar interests and to predict formation of interconnections between them. Therefore, knowledge of link prediction is of vital importance for social network analysis. Link prediction denotes the probability of adding new links to an already-existing network at a future time. By identifying the specific connections (links) that will be established in the near future, formation of new relationships can be determined and predicted in the community structure.
Interaction event network modeling based on temporal point process
Published in IISE Transactions, 2022
Link prediction (Lü and Zhou, 2011) is a task that identifies whether there is a link between two specific nodes on the network. We adopt the Enron email dataset to evaluate this task. The methods compared include DeepWalk, node2vec, LINE (Tang et al., 2015), HTNE, and the proposed TPPN method. For each of these methods, we use 70% of all the emails as positive samples, and randomly sample pairs of nodes without links in a 1:1 ratio as negative samples. A support vector machine model is used to train a classifier for link identification, and the performance of the model is tested on the remaining 30% of links as well as the negative samples in a 1:1 ratio. The learned dimension is also set as d = 128. We record the accuracy and Macro-F1 of the link prediction task for the compared methods. Table 6 shows the link prediction performance of the experiment. It is shown that the proposed TPPN method outperforms the other methods in both accuracy and Macro-F1.
A machine learning approach for predicting hidden links in supply chain with graph neural networks
Published in International Journal of Production Research, 2022
Edward Elson Kosasih, Alexandra Brintrup
Given a graph G(V, E), where V denotes the set of nodes and E links. We assume that there are hidden links that are not shown in E due to data incompleteness. Link prediction is defined as the task of predicting the existence of a link between two nodes (u, v) ∈ V, (u, v) ∉ E. We assume that the graph is undirected. In practice, supply chains are directed, but most link prediction approaches simplify this and convert it into an undirected graph. We stick to this paradigm as the directionality of the links may become obvious to the practitioner once the relationship is known. However, working on the original directed graph would be an interesting future extension.
Towards knowledge graph reasoning for supply chain risk management using graph neural networks
Published in International Journal of Production Research, 2022
Edward Elson Kosasih, Fabrizio Margaroli, Simone Gelli, Ajmal Aziz, Nick Wildgoose, Alexandra Brintrup
As mentioned before, determining the satisfiability of predicate logic statements derived from KG triplets is closely related to another machine learning task called link prediction. The term ‘Link prediction’ is mainly used to refer to predictive tasks in homogeneous graphs, where there is only single entity type and relation type, hence all nodes and edges are of the same types. Link Prediction is used in many domains such as social network analysis, recommendation system or e-commerce. Main types of link prediction algorithms are discussed below.