Research On Theme Mining And Evolution Law Prediction Of Multi Domain Scientific And Technological Resources

There are a lot of researches on the field classification and subject analysis,at present,the data collected by some small websites are complicated and noisy,the data semantics are sparse and there is no subdivision category,and the theme of demand description is not clear,which leads to the inaccurate analysis of the theme development law of science and technology demand,and there is no hot development trend analysis of science and technology demand,resulting in the unclear hot direction of science and technology demand.Therefore,it is very important for researchers to study science and technology demand.Based on neural network,clustering algorithm,convolution network and other technologies,this paper extracts the demand topics in multi domain science and technology needs through the topic extraction model,clusters and divides them according to the extracted topic characteristics,analyzes the evolution law of different dimensions of the demand topics in each industry field according to the obtained topic words and domain classification,and forecasts the development trend of the heat of the topic,The prediction results and the theme heat evolution law of science and technology demand are visually recommended to researchers,and the theme extraction and evolution law prediction system of multi domain science and technology resources is realized.The work completed in this thesisi can be divided into the following four points:(1)Aiming at the problems of short text and sparse semantics of multi domain science and technology resource information,an industry domain classification algorithm based on multi domain science and technology requirements is proposed.This thesis standardizes the data,realizes the text Vectorization Based on word2vec model,and processes the text information.It classifies the scientific and technological demand field,and does not obtain the corresponding label for the scientific and technological demand field.(2)A topic extraction method of multi domain science and technology resources based on lda2vec model is proposed.Taking the industry field of science and technology demand data as the first level subject division,the subjects in each field classification are extracted by combining textrank and lda2vec improved subject extraction method(TEL),and the subjects are determined by combining the importance of vocabulary,semantic relationship and the relationship between the whole context and vocabulary.The obtained theme features are clustered,divided into secondary categories,and the themes of this category are extracted.The secondary theme representatives are determined according to the scores of all themes in the subdivided secondary categories,so as to complete the fine-grained classification and theme acquisition of science and technology needs,and describe the distribution of demand themes in various industries and fields of science and technology needs.(3)A topic heat prediction method of science and technology demand based on time convolution network(TCN)is proposed.According to the obtained topic characteristics of science and technology demand,the topic heat prediction in time sequence is carried out based on TCN network and self attention mechanism.Combined with self attention mechanism technology,the network can notice the correlation between different parts of input,Learning local important information increases the accuracy of topic heat prediction of science and technology demand data.The analysis results of science and technology demand topics and the prediction results of topic evolution law are presented to users.(4)The subject extraction and evolution law prediction system of multi domain science and technology demand is designed and implemented.The function analysis of the algorithm processing results of science and technology demand data in the corresponding application scenarios is made,and the processing results are visualized in the system.The system designs three modules:multi domain science and technology demand topic extraction,multi domain science and technology demand topic evolution law prediction and visualization and multi domain science and technology demand recommendation,and tests the system.This thesis realizes the acquisition and preprocessing of multi domain scientific and technological resources information,the subject extraction of multi domain scientific and technological resources information,the evolution law prediction of multi domain scientific and technological resources data for researchers,and finally realizes the subject extraction and evolution law prediction system of multi domain scientific and technological resources.It can realize the theme distribution of science and technology demand data,fine-grained classification of industry fields,Theme Evolution Law and future prediction results,and recommend the results of science and technology demand analysis to researchers,which has certain practical value.