Causal User Clustering for Accurate KPI Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data processing systems fail to accurately identify user clusters due to ignoring causal effects among sub-groups of users, leading to inaccurate identification of key performance indicators.
Innovation Solution
A machine learning model generates a directed graph representing causal relations among user interactions with a digital platform, allowing for the recognition of heterogeneity among sub-groups and updating user clusters based on these relations, thereby providing customized content with increased accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If homogenous clustering criteria are used to group users, then the clustering process is simple and fast, but the accuracy of identifying key performance indicators deteriorates
Solution Approach 1:
The patent segments users into distinct clusters based on heterogeneous causal relationships rather than uniform similarity criteria. By dividing the user base into sub-groups with different causal structures (e.g., users who respond to content A versus users who respond to content B), the system achieves both manageable cluster sizes for processing and high precision for performance indicator identification within each segment.
Solution Approach 2:
The patent applies local quality by allowing each user cluster to have its own causal relationship structure and performance indicators rather than imposing a single homogenous model on all users. Each cluster is characterized by locally relevant causal factors (e.g., different content types, different interaction patterns), enabling accurate identification of cluster-specific key performance indicators while maintaining overall system efficiency.
2Device complexity
If conventional homogenous clustering is used, then the system complexity is low, but the accuracy of user targeting deteriorates
Solution Approach 1:
The patent introduces dynamics by enabling the causal relationship model to adapt to different user clusters rather than using a static homogenous model. The system dynamically identifies and updates causal relationships specific to each cluster, allowing the complexity to be managed through iterative learning processes while achieving high targeting accuracy through cluster-specific causal models that evolve with data.
Solution Approach 2:
The patent changes parameters by allowing different causal relationship parameters for different user clusters. Instead of using fixed homogenous parameters, the system estimates and updates cluster-specific parameters (e.g., different weightings for various interaction types, different causal graphs) to optimize targeting accuracy for each segment while managing overall system complexity through parameterization techniques.
3Ease of manufacture
If homogenous clustering is applied, then the implementation is straightforward, but the effectiveness of customized content deteriorates
Solution Approach 1:
The patent segments the content delivery approach by creating cluster-specific content strategies rather than using a single homogenous content approach. Each user cluster receives customized content based on its specific causal relationships (e.g., different content types, different messaging approaches), significantly improving the reliability and effectiveness of customized content while maintaining implementation feasibility through modular cluster management.
Solution Approach 2:
The patent applies local quality to content customization by tailoring content characteristics to match the causal relationships of each user cluster. Each cluster receives content with locally optimized properties (e.g., different content formats, different timing strategies) that align with its specific causal structure, dramatically improving content effectiveness while keeping implementation manageable through standardized local quality frameworks.
Data Source
AI summary
Methods, non-transitory computer readable media, apparatuses, and systems for data processing include obtaining, by a machine learning model, a user cluster and interaction data for users in the user cluster, where the interaction data relates to interactions between the users and a digital platform. Some embodiments further include generating, by the machine learning model, a directed graph based on the user cluster and the interaction data, where the directed graph represents causal relationships among the interactions. Some embodiments further include updating, by the machine learning model, the user cluster based on the directed graph. Some embodiments further include providing, by a content component, customized content to a user via the digital platform based on the updated user cluster.