Consumer Web Browsing Prediction Using Weighted Clickstream Link Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies fail to accurately leverage clickstream data for predicting consumer web browsing behaviors, leading to inefficient and costly marketing efforts.
Innovation Solution
A method involving the creation of a link graph using nodes for webpages, links for hyperlinks, and weights for interaction counts, allowing for the determination of a programmatically expected path for consumer browsing, enabling targeted marketing and resource optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional marketing methods are used without accurate consumer behavior prediction, then marketing coverage is broad, but marketing efficiency is low and costs are high
Solution Approach 1:
The system performs preliminary analysis of clickstream data to predict consumer browsing paths before marketing actions are executed. By pre-processing historical data to create link graphs and predict expected paths, the system enables targeted marketing interventions that occur at optimal moments in the consumer journey, thereby improving efficiency and reducing wasted spend on irrelevant marketing impressions
Solution Approach 2:
The system continuously collects clickstream data from user interactions and feeds this information back into the prediction model. This feedback loop allows the system to refine its understanding of consumer behavior patterns over time, improving the accuracy of path predictions and enabling increasingly precise marketing targeting that reduces costs while maintaining high efficiency
2Measurement precision
If detailed clickstream data is collected and analyzed, then prediction accuracy improves, but data processing complexity increases
Solution Approach 1:
The system segments the complex task of behavior prediction into distinct components: data collection, link graph construction, expected path calculation, and prediction generation. By processing clickstream data through these separate stages and maintaining organized data structures (link graphs with nodes and edges), the system achieves high prediction accuracy while managing processing complexity through modular architecture
Solution Approach 2:
The link graph serves as an intermediary data structure that bridges raw clickstream data and prediction outputs. By transforming detailed interaction data into a structured graph representation with webpages as nodes and hyperlinks as edges, the system enables efficient path analysis without requiring direct processing of all raw data points, thus improving accuracy while controlling complexity
3Reliability
If comprehensive historical data is stored for all users, then prediction reliability improves, but storage requirements and processing overhead increase
Solution Approach 1:
The system extracts only the essential elements needed for prediction from comprehensive clickstream data: webpage visit sequences, hyperlink click patterns, and temporal information. By extracting these key features and storing them in link graphs rather than retaining all raw interaction data, the system maintains prediction reliability while significantly reducing storage requirements and processing overhead
Solution Approach 2:
The system performs preliminary aggregation of clickstream data into link graphs that capture essential browsing patterns. By pre-processing and structuring data into compact graph representations that encode user behavior patterns, the system preserves the information necessary for reliable predictions while minimizing the volume of stored data through efficient compression and abstraction
Data Source
AI summary
Embodiments of the present invention provide methods, systems, apparatuses, and computer program products for predicting consumer behavior. In one embodiment a method is provided comprising automatically creating a link graph comprising nodes representing webpages, links representing hyperlinks, and weights for each link representing a number of times a hyperlink associated with the respective link redirected the a user devices from a webpage associated with a first node connected to the link to a webpage associated with a second node connected to the link; and determining based on the link graph a programmatically expected path for a particular user, wherein the programmatically expected path identifies, at least, two or more webpages that the particular user is programmatically expected to visit and specifying a programmatically expected order at which the particular user will visit the webpages.