基于并行K-均值算法的网络广告精确营销系统设计

发布时间：2018-04-22 11:14

本文选题：网络广告 + 数据挖掘　；参考：《山东科技大学》2007年硕士论文

【摘要】： 可定向投放是网络广告的一大特点。数字时代的新技术使网络广告的定向投放成为可能：网络广告可以投放给某些特定的目标人群，甚至可以做到一对一的定向投放。数据挖掘作为一种先进的数据分析方法，，是实现对客户数据进行深入分析的有效工具，其中聚类分析是其中重要的一种作用，可实现在没有训练的条件下把样本分为若干类。网络广告精确营销系统是聚类算法在网络广告投放中的应用体现，帮助用户锁定目标受众进行合适的网络广告投放。本文在研究了网络广告的特点和发展历史，以及数据挖掘的基本知识的知识背景下，提出了细分受众的精确网络广告营销的思路，并在此基础上提出了基于并行K-均值算法的网络广告精确营销系统模型，并详细说明了数据库管理，前端处理系统，数据采集系统，后台数据分析系统和效果计费系统的设计原理，并重点描述了基于并行K-均值算法的受众聚类系统的设计。根据受众的点击率、点击数、转化率，近期访问量以及曝光时间总和五个属性确定受众评价函数，利用PVM搭建并行虚拟机，选用Master／Slave模式实现了细分受众模块的程序设计，并最后通过实际效果证明了本系统的优越性。在本文最后还对系统的进一步改进作了总结，为下一步工作指明了方向。通过并行算法细分受众，快速的锁定目标受众群，实施合适的网络广告营销策略，这是网络广告投放市场的一次尝试，具有重要的现实意义和理论意义。
[Abstract]:Targeted delivery is a major feature of online advertising. The new technology of the digital age makes it possible to direct the network advertisement: the network advertisement can be delivered to some specific target groups, and even can be directed to one to one. As an advanced data analysis method, data mining is an effective tool for in-depth analysis of customer data, among which cluster analysis is an important role, which can be used to classify samples into several classes without training. The network advertisement precise marketing system is the application embodiment of clustering algorithm in the network advertisement putting in, helps the user to lock in the target audience to carry on the suitable network advertisement to put in. Under the background of studying the characteristics and development history of network advertisement and the basic knowledge of data mining, this paper puts forward the idea of accurate network advertising marketing of subdividing audience. On this basis, the model of network advertisement precise marketing system based on parallel K-means algorithm is put forward, and the database management, front-end processing system and data acquisition system are described in detail. The design principle of background data analysis system and effect accounting system is described, and the design of audience clustering system based on parallel K-means algorithm is described. According to the five attributes of audience click rate, click number, conversion rate, recent visit amount and exposure time summation, the audience evaluation function is determined, and the parallel virtual machine is built by using PVM, and the program design of subdivision audience module is realized by using Master/Slave mode. Finally, the superiority of the system is proved by the actual effect. At the end of this paper, the further improvement of the system is summarized, and the direction of the next work is pointed out. Subdividing the audience through parallel algorithm, locking the target audience quickly, and implementing the appropriate network advertising marketing strategy is an attempt to put online advertising into the market, which has important practical and theoretical significance.
【学位授予单位】：山东科技大学
【学位级别】：硕士
【学位授予年份】：2007
【分类号】：TP311.52

【引证文献】