Score Normalization Model for Predictive Output Consistency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Predictive models generated using different training functions produce scores on varying scales, leading to inconsistent outputs, which can disrupt user experience when replacing or updating models.
Innovation Solution
A score normalization model is generated using intermediate training records to normalize scores between zero and one, ensuring consistent predictive outputs across different models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If predictive models are updated or replaced with new training functions, then model performance and accuracy can be improved, but score scales vary between models leading to inconsistent outputs
Solution Approach 1:
A score normalization model is introduced as an intermediary component between the predictive model and the output interface. This normalization model receives scores from different predictive models trained with various training functions and transforms them into a unified score scale, ensuring consistent outputs regardless of which predictive model is used.
Solution Approach 2:
The patent applies parameter changes by transforming the score parameter through a normalization function. The score normalization model adjusts the range and distribution of scores to fit a standard scale, changing the parameter characteristics while preserving the relative ordering and predictive information.
2Adaptability or versatility
If different training functions are used to generate predictive models, then model diversity and optimization opportunities increase, but output consistency decreases
Solution Approach 1:
The score normalization model serves as a mediator that accepts inputs from diverse predictive models trained with different training functions and produces standardized outputs. This intermediary layer decouples the diversity of training approaches from the consistency of outputs, allowing both goals to be achieved simultaneously.
Solution Approach 2:
The score normalization model is designed with universal functionality to handle scores from any predictive model regardless of which training function was used. It provides a unified interface and consistent score distribution that works across multiple model types and training approaches.
3Ease of operation
If score normalization is implemented to maintain consistent outputs, then user experience stability improves, but system complexity increases
Solution Approach 1:
The system is segmented into distinct functional components: the predictive model layer for generating scores and the score normalization layer for standardizing outputs. This segmentation allows each component to specialize in its function, making the overall system more manageable despite the added complexity of normalization.
Solution Approach 2:
By introducing the score normalization model as an intermediary, the patent isolates the complexity of score transformation in a dedicated component. This mediator handles all the complexity of adapting different score distributions, leaving the predictive models themselves simple and the user interface consistent.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for score normalization. One of the methods includes receiving initial training data, the initial training data comprising initial training records, each initial training record identifying input data as input and a category as output. The method includes generating a first trained predictive model using the initial training data and a training function. The method includes generating intermediate training records by inputting input data of the initial training records to a second trained predictive model, the second trained predictive model generated using the training function, each intermediate training record having a score. The method also includes generating a score normalization model using a score normalization training function and the intermediate training records.