Node state prediction method based on a graph convolutional neural network

Abstract

The invention discloses a node state prediction method based on a graph convolutional neural network. Through a graph convolutional neural network algorithm, a network structure of data and a node state at a certain moment can be input, the node state can be a continuous state value or a discrete class state, and then the node state at a next moment is predicted. The node state prediction of the data of the network structure is very significant, and the method can be applied to many fields, such as flow prediction on a traffic network, infection condition prediction on a disease transmission network and the like. A traditional state prediction method generally neglects a network structure to result in inaccurate prediction results, so that the method for predicting the node state based onthe network structure provided by the invention not only has a better effect, but also has a wide application field.

Description

technical field [0001] The node state prediction is performed on the data of the network structure, specifically related to the technical field of the graph convolutional neural network algorithm. Background technique [0002] In the real world, many important data exist in the form of network. Network data is data composed of nodes and edges, representing objects and their connections. Nowadays, the network is one of the most commonly used data types. For example, the road connections between cities constitute a transportation network, the relationship between people constitutes a social network, and the citations between scientific research papers constitute a scientist’s cooperation network, etc. Wait. For different networks, nodes and edges represent different meanings, and the network structure also reflects the mutual influence relationship between nodes. In addition, for network data, in order to fully reflect the information of the data, nodes and edges can also h...

AI Technical Summary

Problems solved by technology

However, for node state prediction of network data, such as traffic flow prediction of traffic network, epidemic prediction of disease transmission network, trade volume prediction of trade network, etc., because of the existence of network structure, traditional methods are difficult to apply directly. When processing data, the underlying network structure is often ignored, and only time-series node status information is considered

However, this approach misses important network structure information, that is, does not consider the mutual influence between nodes, which is bound to be unfavorable for the prediction results.

Images