Graph neural network GAT-based patent valuation method and system for click auction website

Abstract

The invention provides a graph neural network GAT-based patent valuation method and system for a click auction website, and relates to the field of evaluation methods. The weights of the patent attribute parameters in a training data set are trained by using an attention graph neural network GAT, a patent valuation model is abstracted into a knowledge graph, an assessed value of the patent value is taken as a Root of the graph, an attribute parameter related to the patent valuation is taken as a neighbor node of the root, and an edge between the Root and the Node represents a weight coefficient W between the parameter and the assessed value; and the weight coefficient W is trained by using the GAT, after W is trained, a test data set is input, and the evaluation value of each patent data in the test data set can be obtained by multiplying the value of each attribute parameter node in the test data set by the corresponding weight. The weight value is obtained by utilizing big data, andthe value of the patent is integrally evaluated and confirmed, so that the value of patent evaluation is more accurate.

Description

technical field [0001] The invention relates to a patented evaluation method and system based on graph neural network GAT of Dianpai.com, which belongs to the technical field of evaluation methods. Background technique [0002] The patent value analysis index system is a set of index groups that can reflect the overall characteristics of the value of the patent being evaluated, and has internal connections and complementary functions. It is an objective reflection of the intrinsic value of the patent in the transaction. A reasonable and complete index system is a prerequisite for evaluating and analyzing the value of patents. [0003] In Dianpai.com's current patent value evaluation system, the weight ratio of each parameter of the patent valuation model is obtained by analyzing human experience. It is more accurate for a specific field but not applicable to other fields, resulting in the true value of the patent. The difference from the appraised value is too large. In ad...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the deficiencies of the prior art, the present invention provides a patent valuation method and system based on graph neural network GAT of Dianpai.com, which uses GAT to train the weights of each parameter of the valuation model, and solves the problem of incomplete evaluation of existing patents. The problem

Images