Research On Key Technology Forecasting Based On Intelligent Methods

With the booming development of the Internet,systematic breakthroughs in newgeneration information technologies such as big data,artificial intelligence and cloud computing herald a fresh round of industrial revolutions.Digital transformation,mainly characterised by artificial intelligence and big data applications,has become a hotspot to be pursued by various fields.Big data-based decision-making,policy and strategy research is now increasingly being promoted in Europe and the US.The effective governance of an increasingly complex national innovation system urgently needs to be supported by modern scientific methods and tools.Using intelligent methods in this context and accurate forecasting of technological developments can help national and corporate managers to grasp the future direction of technological development,so that they can plan more complete technological development plans,break through blockades in science and technology and gain a helpful position in the future technological competition.This dissertation takes "key technology forecasting" as the research aim,uses papers and patents as data sources,and integrates intelligent methods such as text mining,machine learning and deep learning.It proposes a technology forecasting analysis framework with "technology trend forecasting","technology opportunity forecasting" and "key technology identification" as the major research lines.The main research elements include.(1)Forecasting technology trends based on deep learningThe country's development is now increasingly dependent on technological innovation.Analysing and forecasting the trajectory of technology development and finding the areas of technology with the greatest potential for growth have become a surefire way for governments and companies to achieve technological breakthroughs.This dissertation proposes a deep learning based technology trend forecasting model,which focuses on topic identification and trend forecasting as the entry point of the research.Firstly,based on the Latent Dirichlet Allocation(LDA)model,an LDA-Kmeans++ topic analysis model is constructed to mine the topics of scientific and technical literature from the perspective of semantic analysis.Secondly,a BERTLSTM-based topic identification model is constructed using the natural language processing word embedding method combined with Long Short-Term Memory(LSTM)to categorize the research fields involved in the papers and patents data,to determine the development status of each research field.Finally,an Ensemble Empirical Mode Decomposition(EEMD)combined with deep learning is proposed to forecast the future trend of each research field.The empirical research part of this dissertation takes hightech new energy vehicles as an example.The primary research areas of new energy vehicles are identified and the future development trends of each research area are forecasted.(2)Forecasting technology opportunities based on the GTM-LPIn order to forecast future scenarios of technological evolution and to help firms forecast innovation opportunities for emerging technologies.This dissertation proposes a novel approach to technology opportunity forecasting.Firstly,the text data of the papers and patents are analysed through lexical annotation in natural language processing to obtain key technical words.Secondly,we use Generative Topographic Mapping(GTM)to generate technology maps to identify technology gaps in the paper and patent data,and use GTM inverse mapping to obtain technology combination opportunities.Finally,a deep learning-based Link Prediction(LP)model is constructed using the citation relationships between papers and patents.We analyse the obtained technology portfolio opportunities to achieve forecasting of the development of probability of each technology portfolio.In the empirical research section,lithium-ion battery technology,a popular research area for new energy vehicles identified in the previous section,is used as an example.Technological innovation opportunities in the field of lithium-ion batteries are identified and forecasts are made for each technology opportunity,resulting in technology opportunities with a high probability of development in the field.(3)Key technology identification based on machine learningThis dissertation proposes a method for key technology identification from massive data based on machine learning combined with patent knowledge flow laws.Firstly,the FP-growth algorithm is used to mine the patent data to obtain the popular technologies.Secondly,according to the flow pattern of patent knowledge from the main classification number to the sub-category number,a knowledge flow matrix is constructed and input-output analysis is applied to achieve key technology identification.Finally,the identified key technologies are analysed for technological innovation opportunities,and the layout status of key technologies is studied from both national and enterprise dimensions.In the empirical research section,lithium-ion batteries,a popular research area for new energy vehicles,are still used as an example.The key technologies in the field of lithium-ion batteries are identified,and the current layout of the key technologies is analysed using a Revealed Patent Advantage(RPA)and multiple co-occurrence analysis.This will help enterprises and countries to understand their own technological strengths and weaknesses and layout,identify competitors and select collaborative R&D partners,to break the technological shackles and achieve technological innovation and development.This dissertation combines machine learning,deep learning and other intelligent methods to propose a systematic framework for technology forecasting analysis,which broadens the theory of technology forecasting,enriches the original technology forecasting analysis and research system,and has positive theoretical significance for research in fields related to technology innovation management.In addition,this dissertation uses new energy vehicles as an example to verify the validity of the model,and the analysis results have practical guidance for the development of the new energy vehicle industry.