搜索引擎广告用户行为预测与特征分析

User behavior classification and feature analysis in search advertising

介绍了搜索引擎广告系统的基本运作模式。通过对广告四元组的特征提取、特征值平滑等操作,将广告记录解析成为训练数据,并将数据分为训练集和测试集,使用支持向量机算法并利用训练集训练出的模型将测试集分类,从而预测出用户的行为。通过对特征的分析,得出对用户行为预测准确率影响最大的特征是点击率。实验证明,在使用该模型中所有特征的情况下,分类的准确率能够达到83.17%。

This paper first introduced the basic model of the search engine advertising system （SEAS）, which pointed out us- er clicks were the source of revenue and reveals the importance to predict user behavior. After feature extraction and feature value smoothing on advertising 4-elements, resolved the advertising records into training data and divided then the training data into training set and testing set. It used training set to train the model by SVM and classify the test set to predict user behavior. The evaluation results based on real data show that the model can be used to predict the user behavior in SEAS. Finally, it proves that the most influential feature of user behavior prediction is the click-thought rate and the classification accuracy rate can reach 83.17% in the case of using all features of the model.