關(guān)于單數(shù)據(jù)流和分布式數(shù)據(jù)流挖掘分類算法的研究



摘 要

隨著科學(xué)技術(shù)的發(fā)展，社會(huì)經(jīng)濟(jì)不斷進(jìn)步，在社會(huì)生產(chǎn)的各個(gè)領(lǐng)域中都產(chǎn)生了大量的數(shù)據(jù)，這些數(shù)據(jù)中蘊(yùn)含著大量的豐富的信息。但是，如何處理這些數(shù)據(jù)并從中得到有用的信息，是對(duì)當(dāng)今計(jì)算機(jī)科學(xué)研究的一項(xiàng)重大的挑戰(zhàn)。數(shù)據(jù)挖掘技術(shù)成為了當(dāng)前研究的一項(xiàng)重要的課題。近年來(lái)，單數(shù)據(jù)流的挖掘得到了廣泛的研究，提出了許多有價(jià)值的模型和算法。但是，隨著網(wǎng)絡(luò)環(huán)境應(yīng)用的普及，單一數(shù)據(jù)流的應(yīng)用必然向著多節(jié)點(diǎn)的分布式數(shù)據(jù)流方向轉(zhuǎn)移，并有著廣泛的應(yīng)用前景。本課題闡述了當(dāng)前國(guó)際上關(guān)于單數(shù)據(jù)流和分布式數(shù)據(jù)流挖掘分類算法的研究現(xiàn)狀，按照算法學(xué)習(xí)模式的方法，對(duì)各種分類算法進(jìn)行比較、歸納，同時(shí)，對(duì)分類技術(shù)當(dāng)前所面臨的問(wèn)題和發(fā)展趨勢(shì)進(jìn)行了總結(jié)和展望。在單數(shù)據(jù)流環(huán)境里，增量式學(xué)習(xí)和集成學(xué)習(xí)是兩種典型的學(xué)習(xí)方法；在分布式數(shù)據(jù)流環(huán)境里，集中式挖掘和分布式挖掘是兩種典型的架構(gòu)，各具優(yōu)勢(shì)。

關(guān)鍵字：數(shù)據(jù)挖掘，單數(shù)據(jù)流，分布式數(shù)據(jù)流，Weka

Abstract

With the development of science and technology, as well as the progress of the economics, there are a lot of data in different areas, which contain large amount of information. However, how to handle these data and derive useful information today is such a major challenge of Computer Science. Data mining technology is becoming an important topic in current research. In recent years, the mining of single data stream has been studied extensively and many valuable models and algorithms emerged. But, with the popularity of internet applications, the application of a single data stream towards the inevitable multi-node transfer of distributed data flow direction and has a wide range of applications. This topic describes the current international and distributed on a single data stream of data stream mining Research Classification Algorithm, In accordance with the method of learning algorithms, to compare and to summarized the various classification algorithm, at the same time classification and current problems faced by a summary of trends and prospects. In a single data stream environment, Incremental learning and integrated learning are two typical learning. In a distributed environment where data flow, centralized mining and mining are two typical distributed architecture, they have different advantage.

Keywords: Data Mining ,Single data stream, Distributed data streams, Weka

目 錄

摘 要 1

Abstract 2

第1章 緒　　論 4

1.1本文工作的來(lái)源 4

1.2目的和意義 5

1.3國(guó)內(nèi)外進(jìn)展 5

1.4本文工作的主要內(nèi)容 5

第二章 數(shù)據(jù)流的概述 7

2..1數(shù)據(jù)流管理系統(tǒng)的研究 8

2.2數(shù)據(jù)流在不同領(lǐng)域的應(yīng)用 9

2.2.1在電信數(shù)據(jù)處理方面 9

2.2.2在軍事作戰(zhàn)環(huán)境中 9

2.2.3在科學(xué)計(jì)算領(lǐng)域方面 10

2.3數(shù)據(jù)流的特點(diǎn) 10

2.4數(shù)據(jù)流挖掘框架 11

2.5本章總結(jié) 12

第三章 單數(shù)據(jù)流分類方法研究 13

3.1傳統(tǒng)的分類方法 13

3.2數(shù)據(jù)流分類技術(shù) 15

3.2.1增量式(incremental)算法。 15

3.2.2集合分類器 17

3.3本章總結(jié) 19

第四章 分布式數(shù)據(jù)流分類方法研究 20

4.1分布式數(shù)據(jù)流的定義 20

4.2分布式數(shù)據(jù)流挖掘面臨的挑戰(zhàn) 21

4.3分布式數(shù)據(jù)流相關(guān)系數(shù)計(jì)算 22

4.4基于SPRINT的VHDDS分類方法 23

4.4.1 SPRINT算法簡(jiǎn)介 23

4.4.2 VHDDS分類算法 23

4.4.3算法過(guò)程 24

4.5本章總結(jié) 28

第五章 分析數(shù)據(jù)挖掘工具 29

5.1Weka背景 29

5.2Weka功能 29

5.3Weka的輸入 30

5.4Weka的輸出 31

5.5Weka的可視化 32

5.6本章總結(jié) 32

參考文獻(xiàn) 33

致 謝 35

外文科技資料翻譯 36

英文原文 36

中文譯文 41