Research On The Construction And Visualization Analysis Of Big Data Knowledge Graph For Site Pollution

In recent years,due to the accelerated urbanization process,a large number of rural people have flocked to cities,resulting in a shortage of urban land resources.In order to improve the urban living environment,many cities require the relocation or decommissioning of polluting enterprises in the original urban areas when formulating new development plans.However,this practice actually leaves a large amount of heavily contaminated land within the urban area,causing irreversible effects on the local ecological environment for a long time.Through contaminated site soil contamination information,site activity information,and enterprise production information,site contamination can be traced quickly,effectively,and accurately,so as to obtain the necessary activity information of the contamination source(including pollution-causing behavior,pollution-causing volume,pollution-causing location,etc.)and provide decision support for risk assessment,pollution remediation,and subsequent pollution supervision of site contamination events,and play an It plays an important role in contamination management and pollution prevention and control.However,the existing polluted site survey data presentation and pollution traceability research,the main way to display the survey report,according to the pollutant monitoring data for pollution source tracking.It cannot reflect the connection between the historical pollution situation and the site activities and enterprise activities in the same period,and cannot provide effective guidance suggestions for the subsequent enterprise pollution prevention and control and production activities.Therefore,the use of intuitive and effective data mining methods for pollutant pollution analysis is the main problem currently faced.Therefore,this paper proposes a knowledge graph construction method for multisource heterogeneous data of contaminated sites.According to the different structures of contaminated site data,we use knowledge construction theory,select appropriate entity identification,relationship identification and knowledge fusion methods to extract various types of information of contaminated sites and establish semantic networks to construct knowledge graphs of contaminated site multi-source heterogeneous data,and visualize the correlation paths of knowledge graphs to analyze the pollutants and find the pollutant-causing factors of pollutants.Finally,the WebGIS contaminated site visualization platform is built to visually represent the contaminated site data.The research for this paper covers the following six areas:(1)Research on structured data pre-processing methods: mainly divided into three stages,the first stage is mainly to use the data processing module Pandas and Numpy in python raw data for optimization,mainly including problematic data rejection,data processing,etc.;the second stage is mainly to unify the coordinate benchmark for data from different sources,most of the survey reports around the world use the local coordinate system,mainly Beijing 54 coordinate system,Chongqing coordinate system,etc.,in order to achieve unified management,need to be transformed into a unified coordinate;the third stage of the processed data for library storage,data modeling,refining the table structure,build indexes,create a database according to the data relationship.(2)NLP-based unstructured data preprocessing method is investigated: using a natural language processing based unstructured data preprocessing method,for text data,the main problem is that the format is not uniform and incomplete,which contains a lot of noise,using natural language processing techniques for processing,to derive the high quality word vector data we need,to facilitate the training of deep learning models later.(3)The study constructs a site contamination traceability knowledge ontology model: the site contamination traceability knowledge is defined and decomposed in an object-based dimensional manner to form a unidimensional and all-round knowledge architecture,which serves as a conceptual model for constructing the site contamination traceability model layer.This architecture contains concepts such as soil pollutant,discharge industry and source class for describing the formation and propagation relationship of soil pollutants;it also contains concepts of geographic information class and soil information class for tracing activity processes and pollutant sources.In the process of modeling,we can model the hierarchical and semantic relationships among concepts based on the knowledge architecture of object pollution traceability.Meanwhile,the geometric and spatial relationships in the geographic information of the site are modeled to construct an ontological knowledge model that can be used for the analysis of the contamination-causing factors of the site.It facilitates the analysis of site pollution traceability.(4)The research implements a deep learning model-based knowledge extraction method for contaminated sites: using a deep learning model to extract knowledge from multi-source heterogeneous data through entity identification,relationship identification,and knowledge fusion,and using computer capabilities to automate the extraction of useful knowledge units,including entities,relationships,and attributes needed to build a knowledge graph,thus forming a multi-relationship knowledge network and laying the foundation for the construction of the schema layer of the knowledge graph.(5)The study constructs a knowledge graph graph database: for the structural variability and quantitative complexity of contaminated site information,traditional databases are not good at storing data with large data volume,unstructured and complex relationships,which can easily cause both retrieval and traversal to be extremely difficult and slow,leading to database redundancy.Therefore,Neo4 j graph database,which has advantages in high depth and multi-complex queries,is selected for storing the knowledge graph of contaminated sites.(6)Developed a knowledge mapping visualization platform: the platform is composed of front-end Vue framework + back-end SSM framework,using visualization tools such as Cesium,Echarts and d3.js to visualize and display contaminated sites,and build three-dimensional scenes of sites,and visualize linkage with knowledge mapping.