Machine-Learned Model Scoring Reducing Invocations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine-learned models for predicting online user behavior face challenges due to bias in rules-based models and limited training data, leading to inefficient resource consumption and inaccurate predictions, especially when dealing with unique user interactions and large numbers of entities.
Innovation Solution
A system and method that leverages a machine-learned model by training it on interactions between target and source entities, invoking the model twice with different feature sets to generate accurate scores, reducing the number of model invocations and improving prediction accuracy without increasing computing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the machine-learned model is invoked for each possible user-entity pair to improve prediction accuracy, then prediction accuracy is improved, but computing resources and time consumption increase significantly
Solution Approach 1:
The patent segments the prediction task into two independent model invocations: one for the user and one for the entity. By dividing the user-entity pair prediction into separate user-level and entity-level predictions, the system avoids the computationally intensive approach of invoking the model for every possible pair while still achieving accurate predictions through the combination of these segmented results.
2Productivity
If the model is trained on average user behavior patterns, then the model can make predictions, but it fails to capture unique user intentions and behaviors
Solution Approach 1:
The patent applies local quality by making the model predictions specific to each individual user's characteristics and behavior patterns rather than relying on average user patterns. The model invocation for each user incorporates user-specific features and interaction history, allowing the system to capture unique user intentions while maintaining the ability to generalize from training data.
3Measurement precision
If more training data and feature data are collected to improve model accuracy, then prediction accuracy improves, but the complexity of the system and difficulty of processing increase
Solution Approach 1:
The patent extracts and processes only the most relevant user and entity features separately through two targeted model invocations, rather than attempting to process all possible features and data simultaneously. This extraction approach allows the system to focus computational effort on the most predictive features for each user-entity pair while avoiding the complexity of handling comprehensive feature sets.
Data Source
AI summary
Machine-learned model scoring techniques for reducing model invocations are provided. In one technique, first values for non-pair-specific interaction features that are associated with a first entity are identified. Second values for pair-specific interaction features that are associated with (a) the first entity and (b) a second entity that is different than the first entity are identified. A machine-learned model that has been trained based on the non-pair-specific interaction features and the pair-specific interaction features generates a first score based on the first values. The machine-learned model also generates a second score based on the first values and the second values. A final score for the first entity-second entity pair is computed based on the first score and the second score. Based on the final score, data about the first entity is transmitted over a computer network to a computing device associated with the second entity.