Study On Hotspot Discovery Method For Academic Knowledge Network Based On Multi-Dimensional Heterogeneous Gain Rate

Mining hot topics in academic paper research is an important research direction in intelligence studies,with important theoretical and practical significance.Keywords are highly condensed in academic papers,and they are also common entry points for hotspot discovery.The keyword-based hotspot mining method originated from Zippoff’s law in the 1940 s,forming important method systems such as word frequency statistics,social network analysis,citation analysis,machine learning and other classical keyword evaluation indexes such as keyword word frequency,Centrality,Cited frequency and Page Rank.The classical indexes are widely used because of their intuitive,accurate and efficient features.However,it is found that these metrics generally have problems,including the shackles of power-law distribution that can only mine the general knowledge keywords in the domain,the lack of evaluation metrics from the perspective of global knowledge structure,and the difficulty of sorting out the knowledge structure of the domain to form systematic mining results.This paper proposes a method for mining academic knowledge network hotspots based on a multidimensional heterogeneous gain ratio index system.Firstly,the theoretical idea of measuring the knowledge structure influence of keywords in the academic knowledge network is derived through scientific revolution theory and information gain theory.Then,a heterogeneous gain ratio algorithm is constructed based on the PMI coefficient weighting optimization and network structure entropy,which calculates the structural influence.Secondly,by borrowing the idea of Data Cubes,the K-Shell and Leiden algorithms are introduced to achieve vertical and horizontal cutting of the academic knowledge network,and the heterogeneous gain ratio index is expanded into a multidimensional heterogeneous gain ratio index system,which includes the global heterogeneous gain ratio as the basic index,measuring the influence of keywords on the global structure of the knowledge network;the local heterogeneous gain ratio as the basic index combined with the K-Shell algorithm as the derived index,measuring the influence of keywords on the local structure of the knowledge network;and the word group heterogeneous gain ratio as the basic index combined with the Leiden algorithm as the derived index,measuring the influence of keyword communities on the global structure of the knowledge network.Finally,taking the keywords of 41 core journals in the field of information systems in the past 20 years as the empirical object,an academic keyword co-occurrence network is constructed,and the mining results and hotspot selection thresholds of the multidimensional heterogeneous gain ratio system are calculated and analyzed.Various classic bibliometric indicators are selected as reference objects,and statistical methods such as variance analysis and correlation analysis are used to test and compare the differences and correlations between the multidimensional heterogeneous gain ratio index and the reference index,demonstrating the scientific and effective of the proposed method.The study shows that:(1)The multidimensional heterogeneous gain rate index system is an effective evaluation index to measure the influence of keywords on the knowledge structure of the disciplinary network and to mine the disciplinary research hotspots.Compared with the classical evaluation index,it maintains good relevance,inherits similar valuable information,and identifies the research hotspots keywords with higher academic influence,better knowledge novelty and better stability of hotness.(2)The global heterogeneous gain rate is a bidirectional,structural influence evaluation index,which can uncover both popular research topics and disciplinary classical specialized topics,and is an effective supplement to the existing research hot spot discovery index system,providing a new perspective to evaluate keyword influence from the natural attributes of disciplinary structure.(3)The local heterogeneity gain rate and word group heterogeneity gain rate can mine multidimensional research hot words and hot word groups,reveal the knowledge structure of scientific research fields in a diversified way,reflect the disciplinary characteristics,and effectively expand the means,results and presentation forms of hot word mining.