Lookalike User Identification via LDA Correlation Approximation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current database systems face inefficiencies in identifying user lookalikes due to the resource-intensive process of calculating specific correlations between attributes, which can lead to irrelevant results and biased data sets when new data is introduced, especially in large data sets with thousands or millions of user identifiers and many attribute records.
Innovation Solution
The implementation of a correlation approximation using Linear Discriminant Analysis (LDA) to determine a correlation factor, which adjusts the importance of attributes based on their overlap with user segments, allowing for efficient identification of lookalike users without calculating specific correlations for every attribute, and incorporating a domain adjustment mechanism to handle new data effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If specific correlations between attributes are calculated to identify user lookalikes, then identification accuracy is improved, but computing resources and processing time are excessively consumed
Solution Approach 1:
The patent segments the attribute correlation calculation into two levels: (1) pre-computed segment-level correlations that capture general trends, and (2) on-demand specific correlations only for relevant attribute pairs. This hierarchical segmentation reduces the overall computational burden while maintaining identification accuracy.
Solution Approach 2:
The system performs preliminary calculations of segment-level attribute correlations in advance, storing these results for reuse. When identifying user lookalikes, the system leverages these pre-computed correlations rather than recalculating all correlations from scratch, significantly reducing processing time and resource consumption.
2Reliability
If all attribute correlations are computed for every data update, then data set accuracy is maintained, but processing time and computational overhead increase significantly
Solution Approach 1:
Instead of computing all attribute correlations for every data update, the system performs partial updates by identifying and recalculating only the correlations affected by the new data. This selective approach maintains data set accuracy while minimizing processing time and computational overhead.
Solution Approach 2:
The system implements a feedback mechanism that monitors data changes and triggers correlation recalculations only when necessary. By feedback-driven selective updates, the system maintains reliability of the data set while avoiding unnecessary processing time consumption.
3Adaptability or versatility
If new data is continuously incorporated into user segments, then data freshness is improved, but model stability deteriorates due to frequent updates
Solution Approach 1:
The system adopts a dynamic update strategy where user segment models are updated continuously but selectively. The domain adjustment mechanism dynamically determines whether updates are necessary based on the nature and impact of new data, allowing the model to adapt to fresh data while maintaining stability by avoiding unnecessary updates.
Solution Approach 2:
The system changes parameters selectively rather than updating all model parameters with every new data point. The domain adjustment mechanism identifies which parameters (attribute correlations) need modification and updates only those, preserving model stability while incorporating data freshness where needed.
4Ease of operation
If attribute importance is uniformly weighted, then processing simplicity is maintained, but identification relevance decreases due to ignoring attribute correlations
Solution Approach 1:
The system applies local quality by assigning different importance weights to different attributes based on their correlations with user segments. Rather than uniform weighting, each attribute's importance is locally adjusted according to its specific relationship with the target segment, improving identification relevance while maintaining reasonable processing simplicity through the use of pre-computed correlation factors.
Data Source
AI summary
A database system performs lookalike analysis on a data set including a plurality of user identifiers, which are associated with one or more attribute records. The database system classifies the user identifiers into one or more segments of user identifiers based on the attribute records. The database system performs Linear Discriminant Analysis (LDA) to calculate a measure of importance of the attribute records relative to the one or more segments. The database system auto-correlates the attribute records based on the numbers of attribute records in the user identifier population and the one or more segments. The database system identifies a set of user identifiers relative to one or more segments using the measures of importance and the auto-correlated parameters.