Graph network structure method for urban heterogeneous node classification

Abstract

The invention discloses a graph network structure method for urban heterogeneous node classification, and belongs to the technical field of big data and smart city network construction. An urban heterogeneous graph is created, a graph convolutional neural network and structure information are introduced for training, and the method comprises the steps: acquiring data and label information requiredfor constructing the urban heterogeneous graph, and constructing the urban heterogeneous graph; preprocessing the city heterogeneous graph to obtain a set of input data; constructing a pre-pre-training model according to the obtained input data set; partially migrating the front pre-training model into a rear pre-training model, and constructing the rear pre-training model according to the inputdata set; and integrally migrating the post-pre-training model into the fine adjustment model and constructing the fine adjustment model according to the input data set. According to the method, the problem of insufficient labels in urban event classification is solved, the effect of performing comprehensive event classification by utilizing different types of data sources is achieved, and the method has a wide application prospect in urban event classification.

Description

technical field [0001] The invention relates to a graph network structure method for classifying heterogeneous nodes in cities, and belongs to the technical field of big data and smart city network construction. Background technique [0002] In the construction of a smart city, the establishment of a public service system for event processing is an important part. This system is applied to government platforms such as the mayor's special line to accept various problems reported by citizens. The specific process of the system includes that for the incidents reflected on a daily basis, the relevant personnel first classify the incidents according to the content and nature of the incidents, and then dispatch the incidents to the corresponding departments according to the categories. However, due to the low classification accuracy and low efficiency of this manual classification method, there are still challenges in how to efficiently classify events. [0003] The current auto...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to solve the problem of poor classification accuracy caused by single processing data and insufficient label data in the existing urban event classification system, and proposes a graph network structure method for urban heterogeneous node classification

Images