A Link Prediction Method Based on Deep Dynamic Network Embedded Representation Model

Abstract

The invention discloses a link prediction method based on a deep dynamic network embedded representation model, the method comprising: step 1) constructing a deep dynamic network embedded representation model; step 2) grabbing a large amount of network data from the Internet, and analyzing the network data Carry out preprocessing; Step 3) divide network data into time slices by a certain length of time, and construct network graph G={G under each time slice t‑N ,...,G t‑1 ,G t}, express G with time series adjacency matrix as X={X t‑N ,...,X t‑1 ,X t}; step 4) put {X t‑N ,...,X t‑1} Input deep dynamic network embedding representation model, where {X t‑N ,...,X t‑2} X as a training sample, {X t‑1} is the y value of the training sample, multiple iterations and the stochastic gradient descent method is used to train the deep dynamic network embedding representation model; step 5) move the time window forward by one unit, and set {X t‑N ,...,X t‑1} Input the deep dynamic network embedding representation model, the output is the network connection matrix X at time t t .

Description

technical field [0001] The present invention relates to the field of the Internet, how to use the topological characteristics of the network and the deep learning method in a complex network to carry out embedding learning on the dynamic network, and use the learning results for link prediction, and specifically relate to a deep dynamic network embedding method Represents the link prediction method for the model. Background technique [0002] With the rapid development of Internet and mobile communication technology, people are becoming more and more closely connected. Through the Internet and communication networks, people form a huge complex network. Interaction, communication and influence between people in the network have been integrated into every aspect of life. The research on social network has also gradually attracted attention, and has become one of the research hotspots in the current scientific field. At present, one of the research directions of social netwo...

AI Technical Summary

Problems solved by technology

However, most of these methods use linear models, which have limited ability to capture nonlinear changes, so they cannot model the dynamic network changes at multiple times; in addition, due to the high complexity of the traditional methods, when the network scale is relatively When it is large, it takes a lot of computing resources and time, so it is often impossible to predict all node pairs

Images