Application Research Based On BIRCH Algorithm And Deep Neural Network In Discipline Analysis

The research and analysis of subject hot spots can effectively guide the discipline management and grasp the direction of subject development.However,with the advent of the information age,text data is growing rapidly,and disciplines have also changed in terms of tasks and structure.Facing the challenge of informatization,it is difficult for traditional subject hotspot analysis methods to quickly and accurately process large-scale text data and analyze the direction of subject development.Therefore,in the face of largescale text data,how to solve the problems of traditional hot spot analysis methods and accurately explore the development direction of the subject have important research significance.Based on the current situation of the development of information disciplines and big data technology in a school,this paper improves the calculation method of the distance between CF nodes and the structure of DNN neural network in the BIRCH clustering algorithm based on the research of relevant literature at home and abroad.Analyze and study the development of the information discipline from the macro and micro levels,mainly from the following aspects.First,in the macro analysis of the development of the discipline,by analyzing the clustering process,theoretical basis and application fields of the BIRCH clustering algorithm,the Euclidean distance calculation method is easy to ignore the semantic relationship between text data when calculating the text features of the BIRCH algorithm.Problem,the method of calculating the distance between CF nodes has been redefined by adding cosine distance.At the same time,the clustering process of the BIRCH algorithm is further improved,and the EC-BIRCH algorithm is proposed.Compared with other clustering algorithms,the results show that the EC-BIRCH algorithm improves the accuracy of text classification.Second,in the micro-analysis of discipline development,the DNN deep neural network structure is improved by adding attention layers.The improved DNN neural network emphasizes the proportion of feature words and further extracts feature words with large text contribution rates.The experimental results of multiple text data prove that the deep neural network combined with the attention layer,referred to as the DNN-AF model,can effectively perform deep mining on the text data.Third,the improved two algorithms are applied to the analysis of a school's information discipline development.Based on the data of the discipline development,the EC-BIRCH clustering algorithm is used to make a macro analysis of the main research areas of a school's information discipline development.Based on the macro analysis,the DNN-AF network model is used to deeply explore the future development direction of the discipline.By comparing with the SPSS software multidimensional scale analysis method,the EC-BIRCH algorithm and the DNN-AF network model proposed in this paper are more advantageous than the SPSS software multidimensional scale analysis method in processing large-scale text data in terms of accuracy and effectiveness.