Multi-Model User-Item Interaction Estimation for Recommendation Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current recommendation systems face challenges in efficiently processing user-item interaction data due to high computational requirements and limited data availability, leading to biased content recommendations and reduced user satisfaction.
Innovation Solution
A method and system that progressively pre-selects digital content using distinct estimation models for different types of user-item interaction data, reducing processing power and improving recommendation accuracy by estimating user-item interaction data independently for each type.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single estimation model is used for all user-item interaction data, then the model complexity is low, but the measurement precision of interaction data estimation deteriorates
Solution Approach 1:
The patent segments user-item interaction data into multiple distinct types (e.g., explicit feedback, implicit feedback, different interaction durations). Separate estimation models are then applied to each data type, allowing each model to be optimized for its specific data characteristics. This segmentation resolves the contradiction by improving measurement precision through specialized models while keeping individual model complexities manageable.
Solution Approach 2:
Different estimation models with appropriate complexity levels are assigned to different types of interaction data based on their specific characteristics. For example, simpler models may be used for implicit feedback while more complex models handle explicit feedback. This local quality approach improves overall estimation precision without requiring all models to be uniformly complex.
2Measurement precision
If multiple estimation models are applied to different types of interaction data, then the measurement precision improves, but the device complexity increases
Solution Approach 1:
By segmenting the estimation task into multiple specialized models, each handling a specific data type, the system achieves higher precision for each segment. The overall system complexity increases, but this is justified by the significant improvement in measurement precision for each interaction data type, particularly for sparse or noisy data.
Solution Approach 2:
The patent creates a universal recommendation system framework that can handle multiple types of interaction data through its multi-model architecture. This multi-functional system processes various data types (clicks, views, purchases, explicit ratings) through appropriate specialized models, improving overall system precision while maintaining a unified recommendation output.
3Productivity
If user-item interaction data is limited, then the system operation is simple, but the reliability of content recommendations deteriorates
Solution Approach 1:
When interaction data is limited, the patent applies estimation models that can work effectively with partial data availability. The models are designed to produce reliable estimates even when complete interaction histories are unavailable, using techniques like collaborative filtering and pattern recognition to infer preferences from limited observations.
Solution Approach 2:
The estimation models incorporate feedback mechanisms that continuously improve recommendation reliability over time. By learning from user interactions and adjusting model parameters, the system progressively improves its ability to generate reliable recommendations even with initially limited data, resolving the contradiction between data quantity and recommendation reliability.
Data Source
AI summary
A method and server for estimating interaction data are disclosed. A user and a digital content item that the user has interacted with form an occurred pair. The method includes retrieving (i) interaction data of a first type associated with respective ones of the occurred pairs, and (ii) interaction data of a second type associated with respective ones of the occurred pairs. The first type is distinct from the second type. The method also includes (i) applying a first model to the interaction data of the first type, thereby estimating interaction data of the first type for non-occurred pairs, and (ii) applying a second model to the interaction data of the second type, thereby estimating interaction data of the second type for the non-occurred pairs. The interaction data is estimated such that the non-occurred pairs are associated with respective estimated interaction data of the first type and second type.