Researches On Building Knowledge Graph Of Geographic Information Observation Results And Intelligent Recommendation Method

With the continuous enrichment of earth observation means and the vigorous development of Storage Furniture and Storage Technology,the data scale has increased rapidly every day,and the management for them has been far beyond the reach of human abilities.Therefore,how to remove redundant data from the massive observation data have great significance in the application of GIS.Traditional recommendation system can be used by combining user preference and historical behavior data in massive data by algorithm.However,it has the problem of cold start for the needs of user historical info,and it is difficult to find the logic which behind user using the recommendation content results high repetition rate.Knowledge Graph uses graph structure to express the relationship between concepts or entities,and has good represention and reasoning ability.Therefore,this paper integrates knowledge graph into recommendation system which serve for comprehensive global geographic observation data,and creates personalized and intelligent recommendation service.The main research work of this paper are as follows:(1)Based on expert knowledge,deep learning technology and distributed Internet resource aggregation technology,this paper puts a new method for constructing domain knowledge graph for global comprehensive observation results is proposed.The route has the characteristics of low labor cost,strong replicability and high accuracy of knowledge connection,which can meet the application requirements of knowledge map in different fields.(2)A recommendation scheme which combines knowledge map is proposed.The scheme has the advantages of many traditional methods,one is that it does not need to specify the strategy manually and has high automation.Second,because of the integration of knowledge graph,the recommendation results are intelligent and interpretable.Third,the problem of traditional recommendation algorithm which needs user's historical interactive data which called cold start problem is solved by blending products,users and knowledge graph together.Fourth,have a high level of optimizing of the computing complexity is optimized with high cost and low resource cost.(3)With the flexible combination of cache,load balancing and service degradation,and based on the concept of micro service,this paper puts a recommended service architecture that high concurrency and high availability are given.Compared with the traditional single machine platform application,the service has many advantages,such as better-looking interface,high development efficiency,flexible deployment,high utilization of computing resources,good concurrent support and reliable service.This paper shows that the knowledge extraction which based on Deep Learning can provide a low-cost and rapid construction for Knowledge Grpah.The recommender system service based on Knowledge Graph can achieve good recommendation effect,and improve the intelligence level when satisfying the personalized requirements of recommendation results.The experiment shows that the architecture based on micro service proposed in this paper has good concurrency and availability,and has some pratical value.