【摘要】：空間聚類是空間數(shù)據(jù)挖掘的重要方法之一，有著十分廣泛的應(yīng)用領(lǐng)域。與規(guī)則歸類不同，聚類分析無需背景知識，能直接從空間數(shù)據(jù)庫中發(fā)現(xiàn)有意義的空間聚類結(jié)構(gòu)，有其無可替代的優(yōu)越性。本文在總結(jié)前人工作的基礎(chǔ)上，研究附加限定規(guī)則的空間聚類問題，在普通空間聚類方法的基礎(chǔ)上為其附加空間限定規(guī)則、非空間屬性限定規(guī)則以及方位限定規(guī)則，以使空間聚類更加符合實際應(yīng)用需求。 本文主要內(nèi)容包括： 1．分析了空間數(shù)據(jù)挖掘及空間聚類分析的背景、意義以及相關(guān)理論和技術(shù)基礎(chǔ)。從空間聚類分析的研究背景及意義出發(fā)，分析國內(nèi)外研究進(jìn)展，并指出當(dāng)前研究中存在的主要問題。 2．分析了空間聚類分析和空間分級分析的基本概念，給出了空間聚類分析的基本框架及主要算法，并結(jié)合聚類分析給出了空間分級分析的基本流程,說明了空間聚類分析與空間分級分析結(jié)合使用的實踐意義。分析了限定規(guī)則問題提出的依據(jù)及其必要性，并給出了限定規(guī)則問題的定義及相關(guān)概念。 3．一般情況下，限定規(guī)則可分為兩類，空間限定規(guī)則以及非空間屬性限定規(guī)則。在此基礎(chǔ)上，，本文提出第三種限定規(guī)則，方位限定規(guī)則。三種限定規(guī)則的度量方式不一樣，附加到空間聚類中的方法也不一樣，本文在格網(wǎng)空間中進(jìn)行空間聚類，通過在格網(wǎng)空間中實現(xiàn)三種限定規(guī)則的數(shù)學(xué)運算及轉(zhuǎn)換，將三種限定規(guī)則同時附加到空間聚類上，實現(xiàn)三種限定規(guī)則共同作用下的空間聚類。 4．分級處理與聚類處理一樣都可挖掘出隱藏在數(shù)據(jù)中的潛在規(guī)律，不同的是分級處理需要根據(jù)實際需求設(shè)定分級指標(biāo)及其權(quán)值。本文研究如何依據(jù)限定規(guī)則對空間聚類結(jié)果進(jìn)行分級處理。在傳統(tǒng)的非空間屬性分級基礎(chǔ)上，本文提出了方位因素分級，并給出了其實現(xiàn)方法，使得非空間屬性規(guī)則和方位限定規(guī)則同時參與空間分級分析。此外，對這兩種分級方式進(jìn)行統(tǒng)一的數(shù)學(xué)轉(zhuǎn)換，實現(xiàn)了兩種分級因素同時影響下的空間分級方法。 5．設(shè)計了基于空間聚類的空間數(shù)據(jù)挖掘應(yīng)用實驗系統(tǒng)，實現(xiàn)了附加限定規(guī)則的空間聚類，聚類結(jié)果的分級處理操作。在此基礎(chǔ)上，本文提出了空間聚類一個新的應(yīng)用方向。將空間聚類與空間分級相結(jié)合應(yīng)用于電子地圖的興趣點選擇，并結(jié)合附加空間限定規(guī)則中障礙距離的計算方法，實現(xiàn)了實時交通環(huán)境下的路徑規(guī)劃。
[Abstract]:Spatial clustering is one of the important methods of spatial data mining, which has a wide range of applications. Different from rule classification, clustering analysis can directly find meaningful spatial clustering structure from spatial database without background knowledge, which has irreplaceable advantages. In this paper, on the basis of summarizing the previous work, we study the spatial clustering problem of additional restricted rules. On the basis of ordinary spatial clustering methods, we make use of the additional space rules, non-spatial attribute rules and azimuth rules. In order to make space clustering more in line with the actual application needs. The main contents of this paper are as follows: 1. The background, significance, theoretical and technical basis of spatial data mining and spatial clustering analysis are analyzed. Based on the research background and significance of spatial clustering analysis, this paper analyzes the research progress at home and abroad, and points out the main problems existing in the current research. 2. The basic concepts of spatial clustering analysis and spatial hierarchical analysis are analyzed. In this paper, the basic framework and main algorithms of spatial clustering analysis are given, and the basic flow of spatial hierarchical analysis is given, and the practical significance of combining spatial clustering analysis with spatial hierarchical analysis is explained. This paper analyzes the basis and necessity of the problem of qualified rules, and gives the definition and related concepts of the problem of limited rules. 3. In general, the limited rules can be divided into two categories. Space qualification rule and non-spatial attribute qualification rule. On this basis, this paper proposes a third kind of qualification rule, the azimuth rule. The measurement methods of the three kinds of restricted rules are different, and the methods attached to the spatial clustering are also different. In this paper, the spatial clustering is carried out in the grid space, and the mathematical operation and transformation of the three restricted rules are realized in the grid space. Three kinds of restricted rules are attached to spatial clustering at the same time to realize spatial clustering under the joint action of three kinds of restricted rules. 4. Hierarchical processing and clustering processing can mine the potential laws hidden in the data. The difference is that the grading process needs to set the grading index and its weight value according to the actual demand. In this paper, we study how to classify the spatial clustering results according to the limited rules. Based on the traditional classification of non-spatial attributes, this paper puts forward the classification of azimuth factors, and gives its implementation method, which makes the non-spatial attribute rules and azimuth defining rules participate in the spatial classification analysis at the same time. In addition, through the unified mathematical transformation of these two classification methods, the spatial classification method under the influence of two grading factors is realized. 5. A spatial data mining application experiment system based on spatial clustering is designed. Spatial clustering with additional qualification rules and hierarchical processing of clustering results are implemented. On this basis, this paper proposes a new application direction of spatial clustering. The spatial clustering and spatial classification are applied to the selection of points of interest in electronic maps, and the path planning in real-time traffic environment is realized by combining with the calculation method of obstacle distance in additional space restriction rules.
【學(xué)位授予單位】：解放軍信息工程大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2013
【分類號】：P208;TP311.13

【共引文獻(xiàn)】

相關(guān)博士學(xué)位論文 前1條

1 付強;中國畜養(yǎng)產(chǎn)污綜合區(qū)劃方法研究[D];河南大學(xué);2013年

本文編號：2157653