Research On Prediction Method Of Growth Of Keywords In Scientific Papers ——Taking Artificial Intelligence Field As An Example

In the context of major changes in the current international situation,it is of practical significance and far-reaching influence to focus on future research hotspots and potential research hotspots and forecast research trends in specific science and technology fields.Keywords in scientific papers is after the author or selective when put in storage,can represent the essence of research contents,research methods,so the use of science and technology literature keywords analysis of high frequency words,suddenly emerge in endlessly,the detection of study but high frequency word analysis mostly stop development vein combing,or qualitative forecast research topics.Burst word detection is used to identify word outbreaks that have occurred.Then,how to make full use of key words in scientific literature to accurately predict the future research trend? Based on the actual needs,this paper starts from the fine-grained literature keywords,fully excavates the historical information of keywords,uses machine learning technology to build the high-frequency word prediction model in scientific and technological literature,explores the burst word prediction method,and predicts the future research trend in the field of science and technology.As the research object,this article selects the field of artificial intelligence to literature for the data source of science and technology,use of science and technology literature citations in the data the author,title,abstract,keywords cited and other relevant information,to carry out the high frequency word prediction research,comprehensive comparison and various combination of features and characteristics,all kinds of algorithm and the effect of parameters on the validation set and test set,and finally get the best effect of the model.In Top100,the coincidence rate and ranking correlation between the optimal high frequency word prediction model and the real results reached 0.85 and 0.75,far exceeding the traditional method(coincidence rate and ranking correlation were only 0.77 and0.68),which proved that this model performed well in both high frequency word prediction and high frequency word ranking prediction tasks.Then,based on the accurate prediction of the word frequency ranking of high-frequency words,this paper explores three prediction methods of burst words,including ranking difference method,ranking deceleration method and combination judgment method.The advantages and disadvantages of the three methods are compared,and the results show that the combination judgment method is superior to the other two methods,with both stability and accuracy.Finally,the combined judgment method was compared with the burst word detection function in Cite Space,and it was found that the burst word detection in Cite Space had time lag and was not suitable for the prediction task.The combined judgment method proposed in this paper can predict the keywords in the initial stage of the outbreak.Finally,based on the prediction model of high frequency words and the prediction method of burst words,the prediction of the field of artificial intelligence in 2020 is carried out.From the perspectives of ultra high fequency words,high frequency words and burst words,the possible and potential research hotspots in the field of artificial intelligence in2020 are analyzed.