Multi-scale link prediction method and system based on hierarchical link mode

A link mode and prediction method technology, applied in the field of network topology, can solve problems such as large memory overhead and computational complexity, inability to capture hierarchical structure, lack of data adaptability, etc., to improve link prediction performance, solve network too sparse, The effect of improving prediction performance

Pending Publication Date: 2022-06-24
LANZHOU UNIVERSITY +1
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the method represented by matrix decomposition often brings excessive memory overhead and computational complexity due to matrix storage and calculation.
However, the graph neural network-based method represented by SEAL needs to construct a subgraph for each pair of nodes for binary classification, which will bring the following disadvantages: 1) The construction process of the subgraph is often too time-consuming; link, you need to construct its corresponding subgraph
[0004] However, the method based on random walk also has obvious defects: one is that it cannot capture the hierarchical structure in the network, and the other is that the graph embedding method represented by Node2Vec constructs manual features of edges in an artificially defined way for binary classification. Lack of sufficient adaptability to data

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Multi-scale link prediction method and system based on hierarchical link mode
  • Multi-scale link prediction method and system based on hierarchical link mode
  • Multi-scale link prediction method and system based on hierarchical link mode

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0051] In this embodiment of the present application, drawing on the idea of ​​hierarchical structure modeling and graph sampling, a set of reachability graphs composed of links of different orders between nodes is constructed, and random walks are used to sample the link graphs of different levels. The node sequence is contextually embedded to learn the hierarchical pattern of links and improve the performance of link prediction.

[0052] like figure 1 As shown, it is a schematic flowchart of the multi-scale link prediction method based on the hierarchical link mode according to the first embodiment of the present application.

[0053] First, we extract the k-order link graph between nodes by a hierarchical approach. Secondly, the node sequence is collected by random walk, and the word vector embedding model Word2Vec based on negative sampling is used to process the context information of the node in the walk sequence, so as to obtain the vector representation of the node on...

Embodiment 2

[0076] Based on the prediction model constructed in this application, the second embodiment conducts experiments on eight classical networks and three sparse networks.

[0077] Eight classic network datasets: USAir is an American airline network with 332 nodes and 2126 edges. NS is a scientific research collaboration network for researchers in the field of network science, with 1589 nodes and 2742 edges. PB is an American political blog network with 1222 nodes and 16714 edges. Yeast is a yeast protein interaction network with 2375 nodes and 11693 edges. C.ele is a C. elegans neural network with 297 nodes and 2148 edges. Power is a transmission network in the western United States with 4,941 nodes and 6,594 edges. Router is a router network with 5022 nodes and 6258 edges. E.coli is an interaction network between metabolites of Escherichia coli, with 1805 nodes and 14660 edges.

[0078] Three sparse network datasets: Figeys is a human (Homo sapiens) protein interaction netw...

Embodiment 3

[0107] In the third embodiment, the technical solution of the present application is applied to aviation route planning.

[0108] In this embodiment, taking an air route network composed of existing air routes between several cities in a region as an example, taking cities as nodes, the shortest paths between cities are obtained through breadth-first search, and air routes links with different distance scales are generated based on this. Reachability graph to describe the hierarchical link pattern in the airline network between cities. Then, the node sequence (city sequence) is sampled by random walk, and the context embedding learning of city nodes is performed according to the word embedding method in the language model, and the vector representation of each city under different distance scale link modes is generated. Then, the vector representations of these inter-city routes are merged to obtain the route feature representation for each city. Using cities with existing ai...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a multi-scale link prediction method and system based on a hierarchical link mode, and the method comprises the steps: constructing K k-order link graphs based on a visible graph of a real target network; vector representations of the nodes on different scales are obtained according to the k-order link graph, and the final representation of each node is obtained by splicing the vectors; constructing a dichotomy training set; and establishing a dichotomy model for multi-scale link prediction. The system comprises a link graph module used for constructing K k-order link graphs based on a visible graph of a real target network; the node representation module is used for obtaining vector representations and final representations of nodes on link diagrams with different scales; the classification module is used for constructing a dichotomy training set; and the prediction module is used for establishing a dichotomy model and performing multi-scale link prediction. According to the method, the high-order link of the network can be effectively captured, and the problem that the network is too sparse is better solved; the prejudice of artificially defining edge embedding binary operators is overcome; and the prediction performance is improved.

Description

technical field [0001] The present application belongs to the technical field of network topology, and in particular relates to a multi-scale link prediction method and system based on a hierarchical link pattern. Background technique [0002] The purpose of link prediction is to predict the probability that two nodes in the network are connected, which is crucial for many applications, such as personalized product recommendation in order to improve user experience and business sales in e-commerce networks; online social networking In the network, friend recommendation based on the user's relationship topology; in the biological protein interaction network, it can assist in predicting the interaction between proteins; in the transportation network, according to the existing transportation mode connection between cities, add new traffic line connection. [0003] The development of link prediction can be roughly divided into two stages: in the first stage, a large number of h...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06Q10/04G06K9/62G06N3/04G06N3/08G06Q50/00G06Q50/30G16B15/30
CPCG06Q10/04G06N3/04G06N3/08G06Q10/047G06Q50/01G06Q50/30G16B15/30G06F18/241
Inventor 李龙杰胡江龙陈晓云
Owner LANZHOU UNIVERSITY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap