Integration learning oriented click anti-fraud prediction
- Resource Type
- Conference
- Authors
- Liang, Hua-Xiong; Zhao, Gang
- Source
- 2022 International Conference on Informatics, Networking and Computing (ICINC) ICINC Informatics, Networking and Computing (ICINC), 2022 International Conference on. :246-250 Oct, 2022
- Subject
- Computing and Processing
Learning systems
Transforms
Predictive models
Data models
Fraud
Data mining
Informatics
singular value decomposition
data mining
click-to-counter fraud
integrated learning
feature transform
- Language
Predicting whether an ad click is a normal click or a cheating click is of great importance to digital marketing. In order to solve the problem of information loss in the numerical process of small-scale data in feature engineering, this paper proposes a click anti-fraud prediction method based on SVD (Singular Value Decomposition) and integrated learning in the process of click anti-fraud data exploration. The method uses SVD to mine information on category features in the dataset while increasing the category data dimensionality; Xgboost is used as a feature converter for integrated learning, and the information of each leaf node in the tree is fed into a logistic regression model for click anti-fraud prediction as a feature vector. The experimental results show that both SVD and integrated learning methods can improve the accuracy of click-to-fraud prediction with an accuracy of 89.19%, and to a certain extent avoid the loss of data information in feature engineering.