基于Logistic回归的广告点击行为的影响因素分析

Analysis of the Influencing Factors of Advertising Click Behavior Based on Logistic Regression Method

摘要: 实时竞拍是现代互联网广告行业中非常流行的一种广告投放模式。本研究的主要目的是探索哪些因素是预测广告是否被点击的重要因素。为此，本研究考虑了以下7个指标：平台编码、竞拍低价、是否为全插屏广告、手机运营商、网络状况、设备制造商和广告展现时段。然后建立Logistic回归模型，通过AIC方法做出模型选择。结果显示：1) 竞拍底价越高，广告越不容易被点击。2) 设备制造商：相较于三星、小米、VIVO等设备制造商，苹果点击率较高。3) 全插屏广告，点击率较高。4) 广告展现时段：相较于上午，下午和晚上的点击率较高。

Abstract: Real-time bidding is a very popular model of ad delivery in the modern Internet advertising industry. The main purpose of this study is to explore which factors are important in predicting whether an ad is clicked or not. For this purpose, the following seven metrics were considered in this study: platform code, low bid price, whether it is a full insertion ad, cell phone operator, network condition, device manufacturer and ad presentation time slot. Then a logistic regression model is established and model selection is made by AIC method. The results show that: 1) the higher the bidding reserve price is, the less likely the ad is clicked. 2) Device manufacturers: Apple has a higher click rate compared to Samsung, Xiaomi, VIVO and other device manufacturers. 3) Full insertion ads with higher click-through rate. 4) Advertising display time: Compared to morning, afternoon and evening have higher click-through rates.