【作者】 李志；

【导师】 魏开平；

【作者基本信息】 华中师范大学 ， 计算机应用技术， 2013， 硕士

【摘要】 随着网络技术的飞速发展,互联网广告成为互联网企业最重要的盈利手段之一。越来越多的企业和机构开始研究互联网广告平台,与此同时,很多企业也慢慢地开始从传统媒体广告投放转向互联网广告投放。然而,互联网广告投放的随意性和泛滥性让网民深受其烦,不仅网络广告的投放得不到预期的效果,而且网站点击率也随之下降。针对这种情况,互联网广告的精准投放给互联网广告市场带来了无限生机。精准广告投放即针对用户的个性化向其投放感兴趣的广告,同时真正满足用户对产品需求的信息。目前互联网广告系统中,要做到精准投放主要有三种方式：常见的定向型,主要是针对地理位置、投放时间段等单个属性或者组合属性进行投放；另一种是基于内容的投放方式,这种广告投放系统主要包括提取网页主题词、提取广告文本主题词,计算它们之间的相关性,然后进行广告的投放。而基于用户行为特征的精准广告投放系统主要是在提取到用户的行为特征数据之后,深入挖掘用户的特征数据,然后采用合适的分类算法对用户分类,进而针对用户的特征投放广告。本文通过对互联网广告交易模式的进一步分析,实现了一个互联网广告需求方平台即DSP (Demand Side Platform)原型系统,该系统通过与互联网广告交易平台的对接,主要帮助广告主参与到广告的竞拍中,并且综合用户信息、广告信息等各种信息计算最佳待投放的广告,从而实现广告的精准投放。在用分类算法对用户的特征分类时,常见的分类算法有神经网络分类算法、决策树分类算法及贝叶斯分类算法等,但每种算法都有自己的优缺点,通过对比分析,选择贝叶斯算法作为用户特征分类算法。同时,考虑到每个属性对类属性不同的影响程度,运用信息论的相关知识,设计出改进的贝叶斯算法,经过试验对比,改进的贝叶斯算法比朴素贝叶斯算法的算法分类准确率更高。更多还原

【Abstract】 With the rapid development of network technology, Internet advertising has become one of the most important means to make a profit for Internet companies. More and more enterprises and institutions began to study Internet advertising At the same time, many companies are grudually shifting from traditional media advertising to network advertising. However, the flood of random Internet ads bother Internet users very much, as a result the effect of online advertising is far from people’ expectation and website clicks fall. Given this fact, the precise delivery of Internet ads is so cheering and will bring infinite vitality to the Internet advertising market. Precise delivery of online ads means by taking users’s characteriscs into account, delivering to users the ads to their interest so as to really satisfy their demand for the information of certain product.Currently there are mainly three ways to achieve precise delivery of online ads in the Internet advertising system. Directional delivery, the most common one, means delivering according to the attribute of the location or time period or the combination of both. The second is content-based. Main technology used in this system includes extractiing web page keywords and the advertising text keywords, and then calculating the correlation between them before the delivery. The third one, precise advertising delivery system is based on user’s behavior characteristics. It is realized by digging the user’s behavior after the characteristic data is extracted, and then classifying users according to appropriate algorithm, and at last delivering ads accordingly.Through a profound analysis of the trading patterns of Internet advertising, an Internet advertising demand-side platform DSP (Demand Side Platform) prototype system is created. By docking with the trading platform of Internet advertising, DSP system helps advertisers get involved in the ad auction and by taking factors such as users’ information, ads information, it can calculate the optimum ad to be delivered, thus realizing precise delivery.As for the classification algorithm on the characteristics of the user, neural network classification algorithm, decision tree classification algorithm, Bayesian classification algorithm are most popular. And each has its own advantages and disadvantages. After a comparative analysis, Bayesian algorithm is chosen as the classification algorithm. Whereas considering the various degree of influence of each attribute on the class attribute, an improved Bayesian algorithm is designed by using the knowledge on information theory. The accuracy of this improved algorithm is tested higher than that of the Naive Bayes.更多还原