Search Result Sorting With Grouped Linear Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional online searching techniques fail to provide accurate sorting of search results based on predicted user preferences, often due to implementation difficulties and poor trade-offs between accuracy and complexity, especially when dealing with complex target variable combinations.
Innovation Solution
A computer-implemented method using a machine learning model trained with grouped linear regression to output relevance scores for items based on target variables, allowing for personalized sorting of search results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If gradient boosting machine (GBM) ranker models are used to improve sorting accuracy, then sorting accuracy is improved, but implementation difficulty increases and requires specialized machine learning platforms
Solution Approach 1:
The patent replaces complex, expensive GBM models with simpler, more accessible linear regression models that can be implemented on standard platforms without requiring specialized machine learning infrastructure. This substitution maintains adequate sorting accuracy while dramatically reducing implementation complexity and platform requirements
Solution Approach 2:
The patent modifies the approach by changing from ensemble tree-based methods (GBM) to linear regression with careful feature engineering and regularization. This parameter change in the model architecture allows achieving comparable accuracy on simpler platforms that don't require specialized ML infrastructure
2Ease of manufacture
If conventional sorting techniques are used to simplify implementation, then ease of implementation is improved, but sorting accuracy based on user preferences deteriorates
Solution Approach 1:
The patent performs preliminary feature engineering and data preprocessing to create optimized input features for linear regression. By preparing the data in advance with carefully selected and transformed features, the simple linear model achieves sorting accuracy that would normally require complex models, thus maintaining ease of implementation while improving accuracy
3Ease of operation
If logistic regression models are used to simplify the approach, then ease of implementation is improved, but performance degrades when applied to complex target variable combinations
Solution Approach 1:
The patent segments the target variable into multiple components and uses separate linear regression models for each segment or aspect. This segmentation allows the simple linear regression approach to handle complex target variable combinations by breaking them down into manageable parts, maintaining both ease of implementation and reliable performance
Data Source
AI summary
A method may include receiving data related to a plurality of items and processing the data using a machine learning model. The machine learning model may have been trained to output a score for each of the plurality of items based on one or more target variables and to process the data using a grouped linear regression for groups of items based on sub-divisions of the groups. The method may include storing the output in a data store. Each entry in the data store may include at least an item identifier for an item, a group name, and the score. The method may include receiving search criteria for a search and identifying a set of search results in a group of items. The method may include determining an order of the set of search results and outputting the set of search results.