發(fā)布時間：2018-01-06 23:16

  本文關鍵詞：面向企業(yè)營銷的全景用戶畫像與模型預測　出處：《浙江大學》2017年碩士論文　論文類型：學位論文

  更多相關文章： 全景用戶畫像 迭代決策樹 線性模型 隨機森林 分布式數(shù)據(jù)存儲與管理

【摘要】：隨著網(wǎng)絡和信息技術(shù)的不斷發(fā)展,數(shù)據(jù)在業(yè)務處理基礎上不斷積累,我們從信息技術(shù)時代進入了數(shù)據(jù)技術(shù)時代。企業(yè)營銷方式也從Product,Price,Place,Promotion 這 4P 理論轉(zhuǎn)向了 Consumer,Cost,Convenience,Communication 這 4C理論,以用戶為中心的精準營銷是企業(yè)所需。但是現(xiàn)在的企業(yè)對用戶的認知不清晰,用戶信息不全,為了完善企業(yè)對用戶的認知,本文將研究中心聚焦在全景用戶畫像和模型預測上,并結(jié)合KTV線上到線下的實際場景,最終實現(xiàn)企業(yè)的精細化運營。本文的工作主要包括以下方面:1.本文設計了一套分布式的處理框架。本文用Hadoop分布式文件系統(tǒng)和Hive實現(xiàn)數(shù)據(jù)分布式的存儲和管理;用Impala系統(tǒng)實現(xiàn)用戶畫像的構(gòu)建;用Spark集群實現(xiàn)模型預測;最終實現(xiàn)分布式的數(shù)據(jù)存儲、管理和分析。2.本文實現(xiàn)了基于多源數(shù)據(jù)融合的用戶畫像構(gòu)建。本文從內(nèi)外部數(shù)據(jù)打通,多維度業(yè)務數(shù)據(jù)打通,多方位屬性粒度等特性設計用戶畫像;通過Impala SQL直接獲取、統(tǒng)計變換、自然語言處理、正則匹配、規(guī)則判定、用戶事件模型等方式實現(xiàn)用戶畫像;最終企業(yè)利用用戶畫像實現(xiàn)對用戶的了解,并能夠滿足企業(yè)營銷業(yè)務。3.本文實現(xiàn)了由迭代決策樹和線性模型融合的模型混合方式。該方法利用迭代決策樹實現(xiàn)特征的自動發(fā)現(xiàn),利用樹的路徑擴充特征向量,并結(jié)合線性模型提高模型的精度。本文將該方法應用到用戶性別分類和用戶消費額度預測模型中,并設計多種方案包括隨機森林、迭代決策樹等進行實驗對比,驗證了該方案的有效性和精確性。
[Abstract]:With the continuous development of network and information technology, data is accumulated on the basis of business processing. We have entered the data technology era from the information technology era. The 4P theory of Pricegne place Promotion turns to Convenience. Communication 4C theory, user-centered precision marketing is required by enterprises, but now the enterprise is not clear about users, user information is not complete. In order to perfect the enterprise's cognition to the users, this paper focuses on the panoramic user portrait and model prediction, and combines the actual scene of KTV online to offline. Finally realize the fine operation of the enterprise. The work of this paper mainly includes the following aspects:. 1. This paper designs a set of distributed processing framework. This paper uses Hadoop distributed file system and Hive to realize data distributed storage and management; Using Impala system to realize the construction of user portrait; Using Spark cluster to realize model prediction; Finally, distributed data storage, management and analysis. 2. This paper realizes the construction of user portrait based on multi-source data fusion. Multidirectional attribute granularity and other characteristics design user portrait; The user portrait is realized by Impala SQL, statistical transformation, natural language processing, regular matching, rule determination, user event model and so on. Finally the enterprise uses the user portrait to realize the understanding to the user. And can satisfy the enterprise marketing business. 3. This paper realizes the hybrid model by iterative decision tree and linear model fusion. This method uses iterative decision tree to realize automatic feature discovery. Using the path of tree to expand the eigenvector and combine the linear model to improve the accuracy of the model, this paper applies this method to the user gender classification and user consumption quota prediction model, and designs a variety of schemes, including random forest. The effectiveness and accuracy of the scheme are verified by comparing the iterative decision tree.
【學位授予單位】：浙江大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：F274;TP311.13

【參考文獻】

相關期刊論文 前3條

1 劉海;盧慧;阮金花;田丙強;胡守忠;;基于“用戶畫像”挖掘的精準營銷細分模型研究[J];絲綢;2015年12期

2 周瑞華;;羅伯特·勞特朋 營銷理念的顛覆與重構(gòu)[J];成功營銷;2014年06期

3 伍青生;余穎;鄭興山;;精準營銷的思想和方法[J];市場營銷導刊;2006年05期

，

本文編號：1389958