| The development of Industrial Internet accelerates the process of digital transformation of manufacturing industry,and also accompanies the generation of mass production process data.Mining production data to generate intelligent decision is the key to promote the development of manufacturing intelligence.However,these production data have not been effectively processed and applied in the actual production.In most cases,engineers still need to rely on personal experience to solve the problems encountered in the production process.Knowledge graph is a structured semantic knowledge database,which includes entities,concepts and their relationships in objective world.Under the background of digital era,in view of its advantages in effectively organizing and processing massive information,all walks of life have begun to pay extensive attention to and actively apply this technology in intelligent search,reasoning decision and personalized recommendation,so as to drive intelligent development in the industry.Therefore,it is very important for intelligent manufacturing to establish manufacturing knowledge map and apply it to production practice.In recent years,a large number of scholars have done research on the build and application of knowledge graph,and have achieved good results.However,the existing methods lack visual information hints and interactive feedback optimisation in the process of building knowledge graph,which is not only detrimental to highquality construction of knowledge data,but also further affects advanced application based on knowledge graph.In order to solve the above problems,we proposes a visual analysis-driven interactive knowledge graph construction method to improve the construction quality of knowledge graphs and apply the constructed high-quality graphs to recommendation tasks.Then we come up with a multiple features fusionbased knowledge map-enhanced recommendation algorithm.It alleviates the challenges of data sparse and cold starts faced by traditional collaborative filtering,and enhances the effectiveness of recommendation applications.Our research in this paper is as follow:(1)In response to the problems of low efficiency and difficulty in guaranteeing accuracy in traditional knowledge graph construction methods,this paper designs a knowledge graph entity relationship visual extraction and interactive optimization scheme.Firstly,an interactive error mining method for entity relationship data is proposed,which is guided by semantic clustering and confidence rating to mine suspicious items to help users quickly identify triads with low accuracy;then a semantic similarity-based relationship correction recommendation method is proposed,which is combined with human-computer interaction exploration on the basis of visualisation of the target entity to relationship correction recommendation word cloud to provide powerful support for relationship correction;finally,the above methods are integrated into a visualisation and human-computer interaction system,and the positivity of the scheme in prompting user errors,helping to correct errors,and achieving adaptive optimisation of the model is verified through real-life case studies.(2)We come up with a new algorithm to address the difficulties in poorly recommended due to data sparseness in traditional filtering recommendation methods,which is named Knowledge Graph-Based Multi-Feature Fusion Recommendation(MFFRN).The algorithm uses the constructed high-quality knowledge graph as a source of auxiliary information,and fuses the semantic representation information,attribute and interaction history of knowledge to build personalized recommendations for target population.Finally,our MFFRN model is experimentally validated.The feasibility and effectiveness of our model to enhance the recommendation system presentation is confirmed through comparative tests. |
