The Study on Preventing Click Fraud in Internet Advertising

Zhi Li ； Weichen Jia

《電腦學刊》 31卷3期 (2020/06) Pp. 256-265

https://doi.org/10.3966/199115992020063103020

摘要

Since the 21st century, China's internet technology has continuously developed in stride. It has evolved from a convenient and efficient information transmission tool to a diversified information carrier with large number of users and a network social service application platform, having positive impact on people's life. This study analyzes the current strategies to prevent click fraud, and the implementation process of all strategies is basically the same: firstly, customers are threatened by click fraud, and then they adopt rights protection behaviors, adopt different strategies, investigate the causes of click fraud, and compensate for the losses caused by click fraud. The strategy to prevent click fraud proposed in this paper starts from the effective number of clicks. Even if there is a malicious click, the click behavior will be regarded as invalid, and no record will be made in the database of the service provider, so that customers will not be threatened by click fraud. This paper uses the random forest algorithm in machine learning to classify features to determine fraudulent click behavior. The results show that the proposed method is higher than 91% in the prediction accuracy of positive and negative samples. In addition, in comparison with several other classification methods, the random forest classification algorithm Internet advertising click fraud detection effect is the most effective. The results of CPC advertising show that the algorithm is effective and feasible.

關鍵字

click fraud ； Cost-per-Click (CPC) ； Internet advertising ； strategy