Web Traffic Prediction Using Tag-Based Profile Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Operators of online resources face challenges in distinguishing genuine from invalid user traffic, which is crucial for accurate advertising billing and optimizing website interactions, as existing methods lack effective prediction tools to characterize user behavior based on website features.
Innovation Solution
A computer-implemented method that characterizes web sites using features and assigns initial tags to model user traffic, recursively updating these tags and profiles to generate a predicted traffic profile, allowing for the differentiation of genuine and invalid user activity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional traffic counting methods are used, then traffic volume can be measured, but the ability to distinguish genuine from invalid user activity is lost
Solution Approach 1:
The patent segments the traffic measurement problem by creating separate processing paths for different types of traffic analysis. It divides user activity into multiple dimensions (clickstream data, session characteristics, navigation patterns) and processes each segment independently before integrating results, enabling both volume measurement and authenticity verification simultaneously
Solution Approach 2:
The patent introduces an intermediary traffic prediction model that acts as a mediator between raw traffic data and billing decisions. This model uses trained prediction algorithms to generate expected traffic profiles, which then serve as a reference standard to validate actual traffic patterns, enabling distinction between genuine and invalid activity without compromising measurement capability
2Reliability
If detailed user behavior tracking is implemented, then genuine user activity can be identified, but system complexity and computational resources increase
Solution Approach 1:
The patent applies preliminary action by pre-training traffic prediction models using historical clickstream data before actual traffic validation occurs. The system performs offline training to establish expected traffic patterns and user behavior profiles in advance, so that during operational phases, only lightweight comparison and validation operations are needed, reducing real-time computational complexity
Solution Approach 2:
The system implements self-service by using its own historical traffic data to train and refine its prediction models. The traffic prediction component continuously learns from accumulated clickstream data, automatically improving its ability to distinguish genuine activity without requiring external intervention or manual configuration, thereby managing complexity through self-optimization
3Measurement precision
If traffic prediction models are trained on historical data, then prediction accuracy improves, but training time and computational resources are consumed
Solution Approach 1:
The patent implements continuity of useful action by establishing an ongoing, iterative training process where the traffic prediction model continuously learns from new clickstream data as it accumulates. Rather than periodic batch training, the system maintains continuous learning operations that gradually improve prediction accuracy over time without requiring system downtime or interrupting service operations
Solution Approach 2:
The system applies partial action by implementing incremental training approaches where the model processes and learns from subsets of data over time rather than requiring complete retraining on all historical data. This allows the system to achieve sufficient prediction accuracy through progressive learning, reducing the time and computational resources needed for full model training while maintaining effective prediction capability
Data Source
AI summary
A traffic prediction component may automatically generate predicted traffic profiles for web sites based on tags that characterize the sites. An initial set of tags can be selected for a web site based on a set of predefined rules. An initial traffic profile may be selected based on the initial set of tags. The predicted profile of user traffic is then generated based on the initial set of tags and on the initial traffic profile.