發(fā)布時間：2019-04-28 19:05

【摘要】：隨著近年來互聯(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銷途徑。然而，企業(yè)累計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，僅僅是存儲這些數(shù)據(jù)都是一筆巨大的開銷。由于現(xiàn)有數(shù)據(jù)庫系統(tǒng)的這些不足，導(dǎo)致了企業(yè)空有大量有用數(shù)據(jù)，卻無法提取有用信息的尷尬處境。本文結(jié)合國內(nèi)煙草企業(yè)面對不斷劇增的業(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)處理平臺的可行性，并詳細介紹了Hadoop平臺技術(shù)及其項目結(jié)構(gòu)和體系結(jié)構(gòu)。 為滿足市場需求，首先必須把握市場的實際需求，影響卷煙銷量的市場因素是多樣的。本文基于時間序列分解法預(yù)測模型，建立卷煙銷售預(yù)測模型，并對模型進行了驗證。具體研究內(nèi)容包括以下幾個部分： （1）針對目前煙草企業(yè)全國銷售數(shù)據(jù)來源多、數(shù)據(jù)規(guī)模龐大等特點，，且基于企業(yè)現(xiàn)有數(shù)據(jù)庫的實際情況，分析構(gòu)建數(shù)據(jù)庫營銷系統(tǒng)的必要性，然后對該系統(tǒng)的總體設(shè)計目標(biāo)和模塊進行說明。 （2）分析研究福建中煙營銷平臺的目前狀況，依據(jù)實際需求，著重分析Hadoop在企業(yè)實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平臺的的可行性。針對Hadoop平臺中的關(guān)鍵技術(shù)HDFS和MapReduce做了深入研究，并以實例說明。 （3）在分析Hadoop平臺的的可行性之后，對各省市各規(guī)格卷煙到日的銷售數(shù)據(jù)進行處理，建立銷量預(yù)測模型，考慮到卷煙市場具有季節(jié)周期變化趨勢和長期增長趨勢的特點，建立符合卷煙市場特征的時間序列銷量預(yù)測模型。該預(yù)測模型已經(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)位級別】：碩士
【學(xué)位授予年份】：2015
【分類號】：TP311.13

【參考文獻】

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)庫營銷的應(yīng)用研究[J];辦公自動化;2011年10期

2 程瑩;張云勇;徐雷;房秉毅;;基于Hadoop及關(guān)系型數(shù)據(jù)庫的海量數(shù)據(jù)分析研究[J];電信科學(xué);2010年11期

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

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

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

6 熊慶華;;沒有硝煙的數(shù)據(jù)庫營銷[J];中國信用卡;2008年04期

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

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

9 吳岳忠;周訓(xùn)志;;面向Hadoop的云計算核心技術(shù)分析[J];湖南工業(yè)大學(xué)學(xué)報;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輸需求預(yù)測研究[D];大連海事大學(xué);2011年

本文編號：2467889