Web Page Advertising Identification Using CNN and XGBoost
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for identifying and blocking internet advertising are inadequate, as they rely on limited data sources like URL analysis and lack accuracy in distinguishing between banner and native advertising, and do not provide a comprehensive approach for filtering web page elements.
Innovation Solution
A method utilizing a modified convolutional neural network (CNN) for image classification, combined with URL and HTML classifiers, to identify and classify web page elements as advertising or non-advertising, employing machine learning techniques like XGBoost and Naive Bayes classifiers to determine probabilities of advertising content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple classification models (CNN, XGBoost, Naive Bayes) are combined for comprehensive ad identification, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system divides the advertising identification task into three specialized components: CNN for image-based ad detection, XGBoost for URL analysis, and Naive Bayes for HTML structure examination. Each model focuses on specific features, improving overall accuracy while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The patent combines multiple classification models (CNN, XGBoost, Naive Bayes) into a unified system that processes different aspects of web page elements simultaneously. The integration of these diverse models leverages their complementary strengths to achieve superior advertising identification accuracy.
2Measurement precision
If comprehensive web page element analysis is performed using multiple classifiers, then advertising identification accuracy is improved, but processing time increases
Solution Approach 1:
The system segments the analysis process into parallel tracks: image processing via CNN, URL analysis via XGBoost, and HTML structure examination via Naive Bayes. This segmentation allows simultaneous processing of different feature types, reducing overall processing time while maintaining comprehensive analysis.
Solution Approach 2:
The system applies different levels of analysis depth to different elements based on their advertising likelihood. High-probability advertising elements receive full multi-model analysis, while low-probability elements may receive simplified processing, optimizing the balance between accuracy and processing speed.
Data Source
AI summary
A system of the present invention allows to determine the likelihood that a certain block of a web page is an advertising, such as a banner or a native advertising. The system uses combined and holistic approach to the analysis of units for the purpose of finding the advertising features (links, images and HTML code). The system does not impose any restrictions on possible ways of using it--it only provides tools for determining the probabilities of the unit belonging to the advertising or non-advertising class.