Spatiotemporal Data Mining Of Ecological Indices Based On The DBSCAN Optimization Algorithm

The ecological environment where human beings live is a huge treasure house of data.The change of the ecological environment is a long process.During this process,the ecological data generated by human-environment interaction are constantly updated and accumulated every day.To better use the massive and complex ecological data,it is necessary to find efficient data processing methods.Spatiotemporal data mining can extract potential and useful information from high-dimensional spatiotemporal data,and the DBSCAN algorithm is a common algorithm for spatiotemporal data mining.However,the current DBSCAN algorithm still has problems such as subjective threshold setting and difficulty in expanding the data dimension when mining geological elements and ecological indices.This paper combines GIS and spatiotemporal data mining methods,and uses four representative indicators of ecological environment,namely greenness,humidity,dryness,and heat,for spatiotemporal data mining.While optimizing DBSCAN algorithm,this paper improves the feasibility of the application of GIS and spatiotemporal data mining technology in the ecological field.The main research results are as follows:Firstly,aiming at the defect that the traditional DBSCAN algorithm needs to set the threshold artificially,a method based on parameter optimization strategy and according to the distribution characteristics of spatiotemporal data set was explored.Through research and analysis,the algorithm can provide users with a reasonable threshold reference,with high effectiveness and usability.Secondly,for the problem that traditional DBSCAN algorithms are mostly used to process two-dimensional data,the time dimension was extended,and the data layers of consecutive years were superimposed in the time dimension to form a temporal and spatial three-dimensional trend map of the ecological index.On this basis,a new multi-element hybrid spatiotemporal clustering algorithm was proposed for more than three-dimensional spatiotemporal data clustering.By introducing the histogram to calculate the similarity between multi-attribute clusters,the graph model and the shortest path algorithm to link the most similar clusters,and the traditional DBSCAN clustering algorithm was extended to more dimensions of clustering analysis.Through research and analysis,the algorithm can cluster long-time series elements with different attributes,which has high feasibility and practicability.Thirdly,using Python language,the optimized DBSCAN clustering algorithm was developed into Arc Map add-in in the Arc Py module built in Arc GIS.Through case analysis,spatiotemporal data sets were constructed by four ecological elements of NDVI(greenness),WET(humidity),NDSI(dryness),and LST(heat)in Xishuangbanna to complete the model building and realize the extended application of DBSCAN optimization algorithm in the temporal and spatial variation law of ecological indices.The research deeply explores the spatiotemporal data mining technology in the ecological field based on GIS technology.Through the optimized DBSCAN clustering algorithm,the spatiotemporal distribution characteristics and evolution laws of the four major ecological indices in Xishuangbanna from 2000 to 2020 were explored.The research results can provide a reference for using DBSCAN algorithm to carry out relevant research in Geoscience,ecological environment and other fields,and provide a scientific basis for ecological protection in the study area.