Industrial research report generation method based on multi-source heterogeneous data fusion

Abstract

The invention relates to the field of multi-source heterogeneous data fusion, in particular to an industry research report generation method based on multi-source heterogeneous data fusion, which comprises the following steps: performing task decomposition on an input research report subject based on a selected industry analysis library to generate research report analysis sub-tasks; enabling the multi-dimensional vector of the research report analysis subtask to pass through a first multi-layer perceptron model to generate a task feature embedding vector; enabling the capability vector of the candidate model to pass through a second multi-layer perceptron model to generate a model capability embedding vector; and performing similarity calculation on the task feature embedded vector and the model capability embedded vector to generate a comprehensive similarity, constructing and solving a matching optimization model to determine a matching model in the candidate models, and analyzing the multi-source heterogeneous research report data to generate an industry research report. According to the method, the industry research report is generated by utilizing the advantages of the multi-term model, and the evolution of industry research report generation from single-point intelligence to ecological intelligence is promoted.

Images