【摘要】：隨著近年來(lái)互聯(lián)網(wǎng)的發(fā)展，企業(yè)已經(jīng)擁有龐大的客戶信息數(shù)據(jù)，這些數(shù)據(jù)的積累為企業(yè)提供了一種有效的營(yíng)銷途徑。然而，企業(yè)累計(jì)的客戶信息是非常龐大的，最初搭建的硬件設(shè)備不可能具備處理如此之大的海量數(shù)據(jù)的能力，僅僅是存儲(chǔ)這些數(shù)據(jù)都是一筆巨大的開(kāi)銷。由于現(xiàn)有數(shù)據(jù)庫(kù)系統(tǒng)的這些不足，導(dǎo)致了企業(yè)空有大量有用數(shù)據(jù)，卻無(wú)法提取有用信息的尷尬處境。本文結(jié)合國(guó)內(nèi)煙草企業(yè)面對(duì)不斷劇增的業(yè)務(wù)數(shù)據(jù)，而現(xiàn)有的業(yè)務(wù)數(shù)據(jù)處理能力明顯不足的現(xiàn)狀，分析煙草企業(yè)構(gòu)建Hadoop分布式數(shù)據(jù)處理平臺(tái)的可行性，并詳細(xì)介紹了Hadoop平臺(tái)技術(shù)及其項(xiàng)目結(jié)構(gòu)和體系結(jié)構(gòu)。 為滿足市場(chǎng)需求，首先必須把握市場(chǎng)的實(shí)際需求，影響卷煙銷量的市場(chǎng)因素是多樣的。本文基于時(shí)間序列分解法預(yù)測(cè)模型，建立卷煙銷售預(yù)測(cè)模型，并對(duì)模型進(jìn)行了驗(yàn)證。具體研究?jī)?nèi)容包括以下幾個(gè)部分： （1）針對(duì)目前煙草企業(yè)全國(guó)銷售數(shù)據(jù)來(lái)源多、數(shù)據(jù)規(guī)模龐大等特點(diǎn)，，且基于企業(yè)現(xiàn)有數(shù)據(jù)庫(kù)的實(shí)際情況，分析構(gòu)建數(shù)據(jù)庫(kù)營(yíng)銷系統(tǒng)的必要性，然后對(duì)該系統(tǒng)的總體設(shè)計(jì)目標(biāo)和模塊進(jìn)行說(shuō)明。 （2）分析研究福建中煙營(yíng)銷平臺(tái)的目前狀況，依據(jù)實(shí)際需求，著重分析Hadoop在企業(yè)實(shí)際需求中可以勝任的數(shù)據(jù)處理技術(shù)，分析在煙草企業(yè)現(xiàn)有軟硬件基礎(chǔ)上構(gòu)建Hadoop平臺(tái)的的可行性。針對(duì)Hadoop平臺(tái)中的關(guān)鍵技術(shù)HDFS和MapReduce做了深入研究，并以實(shí)例說(shuō)明。 （3）在分析Hadoop平臺(tái)的的可行性之后，對(duì)各省市各規(guī)格卷煙到日的銷售數(shù)據(jù)進(jìn)行處理，建立銷量預(yù)測(cè)模型，考慮到卷煙市場(chǎng)具有季節(jié)周期變化趨勢(shì)和長(zhǎng)期增長(zhǎng)趨勢(shì)的特點(diǎn)，建立符合卷煙市場(chǎng)特征的時(shí)間序列銷量預(yù)測(cè)模型。該預(yù)測(cè)模型已經(jīng)在企業(yè)中得到應(yīng)用，指導(dǎo)企業(yè)生產(chǎn)和銷售。
[Abstract]:With the development of Internet in recent years, the enterprise already has the huge customer information data, these data accumulation has provided a kind of effective marketing way for the enterprise. However, the accumulated customer information is very large, the initial hardware can not have the ability to deal with such a large amount of data, just to store the data is a huge overhead. Because of the deficiency of the existing database system, the enterprise has a lot of useful data, but can not extract the useful information. This paper analyzes the feasibility of constructing Hadoop distributed data processing platform for tobacco enterprises based on the situation that domestic tobacco enterprises face the increasing business data and the existing business data processing ability is obviously insufficient, and the feasibility of constructing the distributed data processing platform for tobacco enterprises is analyzed in this paper. The technology of Hadoop platform and its project structure and architecture are introduced in detail. In order to meet the market demand, first of all, we must grasp the actual demand of the market, and the market factors that affect cigarette sales are diverse. Based on the prediction model of time series decomposition, this paper sets up a model of cigarette sales forecast and validates the model. The specific research contents include the following parts: (1) in view of the current national sales data of tobacco enterprises from a large number of sources, data scale and other characteristics, and based on the actual situation of the existing database of the enterprise, The necessity of building a database marketing system is analyzed, and then the overall design objectives and modules of the system are explained. (2) analyze and study the current situation of Fujian Tobacco Marketing platform, according to the actual demand, emphatically analyze the data processing technology that Hadoop can be competent in the actual needs of enterprises. The feasibility of constructing Hadoop platform on the basis of existing hardware and software in tobacco enterprises is analyzed. The key technologies of Hadoop platform, HDFS and MapReduce, are studied in depth, and an example is given. (3) after analyzing the feasibility of Hadoop platform, the sales data of every province and city are processed, and the sales forecast model is established. Considering the characteristics of seasonal cycle and long-term growth trend of cigarette market, the cigarette market has the characteristics of seasonal cycle trend and long-term growth trend. The model of time series sales forecast is established which accords with the characteristics of cigarette market. The prediction model has been applied in enterprises to guide the production and sales of enterprises.
【學(xué)位授予單位】：浙江理工大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2015
【分類號(hào)】：TP311.13

【參考文獻(xiàn)】

相關(guān)期刊論文 前9條

1 楊浩;;基于網(wǎng)絡(luò)環(huán)境中數(shù)據(jù)庫(kù)營(yíng)銷的應(yīng)用研究[J];辦公自動(dòng)化;2011年10期

2 程瑩;張?jiān)朴?徐雷;房秉毅;;基于Hadoop及關(guān)系型數(shù)據(jù)庫(kù)的海量數(shù)據(jù)分析研究[J];電信科學(xué);2010年11期

3 向美英;何利力;;趨勢(shì)比率模型在卷煙預(yù)測(cè)中的應(yīng)用[J];工業(yè)控制計(jì)算機(jī);2011年09期

4 李玉林;董晶;;基于Hadoop的MapReduce模型的研究與改進(jìn)[J];計(jì)算機(jī)工程與設(shè)計(jì);2012年08期

5 謝桂蘭;羅省賢;;基于Hadoop MapReduce模型的應(yīng)用研究[J];微型機(jī)與應(yīng)用;2010年08期

6 熊慶華;;沒(méi)有硝煙的數(shù)據(jù)庫(kù)營(yíng)銷[J];中國(guó)信用卡;2008年04期

7 邢慧嫻,楊維中,王漢章;傳染病預(yù)測(cè)[J];預(yù)防醫(yī)學(xué)情報(bào)雜志;2004年06期

8 孔慶凱;;平均預(yù)測(cè)法的應(yīng)用條件[J];預(yù)測(cè);1985年05期

9 吳岳忠;周訓(xùn)志;;面向Hadoop的云計(jì)算核心技術(shù)分析[J];湖南工業(yè)大學(xué)學(xué)報(bào);2013年01期

相關(guān)博士學(xué)位論文 前1條

1 李楠;區(qū)域交通信息集成與運(yùn)輸需求預(yù)測(cè)研究[D];大連海事大學(xué);2011年

本文編號(hào)：2467889