Research On Technology Forecasting Method From The Perspective Of Data Fusion

The advent of the era of big data has brought new challenges to technology management.Traditional technology management methods and tools are no longer able to meet the technology management needs of big data environment.The emergence and development of technology forecasting provides a viable solution to current technology management problems.The current research of technology forecasting has the following problems:(1)Single data source cannot meet the analytical requirements of technology forecasting;(2)The current solution to the multi-source data problem in technology forecasting is not effective;(3)The current data fusion method is not suitable for solving data fusion problem in technology forecasting;(4)In current research,data fusion`s role often stays at a certain stage of scientific analysis,such as data processing,data analysis,currently rare to be able to integrate data fusion into the entire process of science research.In order to solve the above problems,based on the technology forecasting related theory,natural language processing theory and topic model theory,this research conducted researches of data fusion,topic analysis and visualization to build a new technology forecasting method and set the field of antidepressants as the target to apply an empirical study and finally got the following conclusions:(1)From the perspective of data fusion,this research optimizes the main links of technology forecasting method.The research content includes data fusion research,topic recognition research,and visualization research in technology forecasting.Finally,a technology forecasting method from the perspective of data fusion is constructed.And empirical research was carried out in the field of antidepressants.The results prove that this method can solve the problem of multi-source data in technology forecasting,make effective use of multi-source data,and better support technology research and development decision-making activities in the target area.(2)This study analyzes the characteristics of data related to technology forecasting,selects natural language processing technology to solve the data fusion problem in technology forecasting,abstracts the data fusion problem into a text classification problem,and uses natural language pre-train model to build a data fusion model.The method was verified,and the results proved that the constructed data fusion model performed well,indicating the effectiveness of the constructed data fusion model.(3)Based on the characteristics of the fused data format,this study selects the LDA topic model and the LDA2 vec topic model as alternatives to the topic recognition method,and selects the optimal topic recognition method from them.According to the results of empirical research,it is found that the LDA2 vec model is more suitable for topic recognition of the fused data format;the LDA model is more suitable for traditional data formats.At the same time,based on this result,a answer to the cause of the application paradox of the LDA2 vec model was proposed.(4)This study applies the text similarity theory and text similarity calculation method to the analysis of the evolution relationship of topics,and builds an automated technology forecasting visualization process.According to the empirical results,it is found that the results are highly automated and easy to operate.At the same time,they can effectively describe the current state of technology research and development in the target area and predict future development trends.The main innovations of this research include:(1)In view of the data fusion problem in technology forecasting,the data fusion problem in technology forecasting is understood and abstracted from the perspective of natural language processing,and the data fusion problem is transformed into the text classification problem in natural language processing.Based on this,a data fusion model and data fusion process for technology forecasting are constructed.(2)For the topic recognition step in technology forecasting,the current mainstream topic recognition methods,LDA model and LDA2 vec model are compared and researched,and the data types applicable to the two are specified.Aiming at the fused data format,LDAvis was selected as the comparison method,and topic significance and topic distance were used as evaluation indicators.The optimal topic model was selected.(3)For the visualization step in technology forecasting,the technology and methods of topic model and text similarity calculation are used to solve the problem of topic recognition and visualization,and the technology forecasting analysis is automated.