Model Relevance Prediction for Multi-Model Output Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
When using multiple models to predict an objective variable, it is challenging to optimally select or combine their outputs effectively, as users are overwhelmed by the sheer volume of results without clear guidance on relevance.
Innovation Solution
A computer-implemented method calculates the relevance of each model's output to the objective variable based on past values and similarities between current and past timings, allowing for the prediction of the objective variable by weighting similarities according to timing differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If outputs of all models are provided to the user, then complete information is available, but the user becomes confused and unable to make decisions
Solution Approach 1:
The patent introduces a relevance calculation mechanism as an intermediary between the multiple model outputs and the user. This mediator computes relevance scores that quantify how important each model's output is for predicting the objective variable, thereby filtering and organizing the information before presenting it to the user. This resolves the contradiction by maintaining information completeness while improving ease of operation through structured presentation.
Solution Approach 2:
The patent transforms the raw model outputs into a new parameter space by calculating relevance scores and similarities. Instead of presenting raw predictions directly, the system computes similarity metrics between current and past timings, and relevance metrics that weight different model outputs. This parameter transformation makes the information more manageable and interpretable for users.
2Reliability
If multiple models are used to predict the objective variable, then prediction accuracy can be improved, but the complexity of selecting or combining outputs increases
Solution Approach 1:
The patent resolves the complexity issue by introducing automated parameter calculations. The system computes similarity scores between current and historical timings, and relevance scores for each model output, using these parameters to automatically weight and combine predictions. This transforms the complex manual selection process into an automated parameter-driven system that maintains high prediction accuracy while reducing operational complexity.
Solution Approach 2:
The patent implements a feedback mechanism where the relevance calculation uses historical prediction accuracy and similarity metrics to dynamically adjust the weighting of different model outputs. The system continuously learns from past performance, adjusting relevance scores based on which models have been most accurate in similar situations, thereby optimizing prediction accuracy while automating the selection process.
3Measurement precision
If relevance calculation based on past data is performed, then prediction quality improves, but computational time and resources increase
Solution Approach 1:
The patent applies partial action by selecting and focusing computation on the most relevant historical timings and models. Instead of uniformly processing all past data, the system calculates similarities and relevances selectively, prioritizing data points that have higher similarity to the current timing or have demonstrated higher relevance in past predictions. This reduces computational burden while maintaining prediction quality.
Solution Approach 2:
The patent performs preliminary calculations of similarity metrics and relevance scores before the actual prediction is needed. By pre-computing these parameters and storing them for later retrieval, the system reduces the computational burden during real-time prediction. This preliminary action allows the system to maintain high prediction quality while reducing real-time computation time.
Data Source
AI summary
It is preferable to predict an objective variable by optimally selecting or combining the output of a plurality of models. A computer-implemented method is provided that calculates, for each of a plurality of models, a relevance of an output of the model with respect to a value of an objective variable based on the value of the objective variable and the output of the model in the past. The method also calculates, for each of the plurality of models, similarities between a current timing and a plurality of past timings based on the output of the model at the current timing, the output of the model at the plurality of past timings, and the relevance. Additionally, the method predicts the value of the objective variable at a target timing based on the similarities.