Topic Mining And Information Recommendation For Multi-source Heterogeneous Data

With the popularity of digitization and networking,most of the information and intelligence related to enterprises are multi-source heterogeneous data with diversified sources and diverse structures.The scientific,accurate and timely identification and tracking of the frontiers of science and technology from these multi-source heterogeneous data can provide enterprises with timely and effective information services.Based on this,this paper collects the relevant data from different sources in the wire and cable industry,and takes the LDA theme model as the research tool to extract the theme,analyze the intensity of the topic and analyze the evolution of the topic.To grasp the forward problems and trends of wire and cable industry development,it is of high reference value for wire and cable enterprises to understand the market situation and make strategic plans.This paper mainly studies the following aspects:First of all,using the LDA model respectively to different sources of data in the wire and cable industry at home and abroad(including journal articles,conference papers and patents)shall be carried out in accordance with the time slice subject extraction,access and the theme of the term is associated with wire and cable.The results show that the data are similar at home and abroad in subject classification.In terms of content,China focuses on basic practice,while foreign countries focus on scientific research,which is more innovative than domestic ones.Topics of news data extraction mainly include industry inspection and products and technologies of interest within the industry.Secondly,in order to obtain the topic with stronger ability to describe the text content,this paper analyzes the theme intensity based on the result of topic extraction.By calculating the intensity of the theme and according to its general trend,the theme is divided into hot theme,common theme and cold topic.lt is found that the hot topics in China have gradually changed from the material and solid properties of wires and cables to the application effects,and the common themes are mostly the performance,material and experience exchange of wire and cable.Topics related to emerging technologies have only recently emerged;cold-door topics relate to wire and cable materials,machinery,properties and mechanisms.In foreign countries,the flame retardancy of wire and cable is a hot topic;before 2005,the common theme is mainly material and performance,and after 2005,the application of wire and cable in other fields is the main;the theme of cold door is similar to that of our country.From the analysis of news data,it is found that the topic of industry regulation belongs to the hot topic,and the related technology of wire and cable belongs to the common theme.Thirdly,subject strength analysis can only highlight the static characteristics of the subject.In order to further reveal the dynamic development process of the theme,this paper carries out the theme evolution analysis.Research findings:technical methods,products and other topics have important reference value in China.Institutional,material,performance and conference themes need to be combined with other categories of themes to play a more important role.In foreign countries,products,technology,performance,and other topics is of high reference value,machinery and materials class contains part of the need to focus on the theme,but the general reference value is not high.Finally,according to the results of topic extraction,theme intensity analysis and theme evolution analysis,there is a division between "theme of no value","theme of value" and"theme with prospect of development".Combining the specific word items included in the topic,the topic is analyzed to determine the content of information recommendation.From the research results,using LDA topic model for topic mining of multi-source and heterogeneous data can obtain topics with clear semantics and strong text description ability.The results of thematic analysis can not only reflect the market situation,but also can be cross-linked with other fields involved in the topic to form a semantic knowledge network.Thus,the validity of using the LDA theme model as a service object to mine the multi-source heterogeneous data and recommend the information is validated.