Data Transaction Profile Expansion for Rapid Scoring Maturation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data transaction scoring systems face challenges in improving performance with increased data feeds due to lengthy maturation times and operational impacts, such as resource consumption and the need for continuous parameter modifications during model evolution.
Innovation Solution
Introducing a first set of variables with lower maturity levels into the system by expanding the data transaction profile, assigning initial values based on a statistical model, particularly using population means, to enable immediate scoring and rapid maturation without operational disruptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of data feeds (variables) is increased to improve scoring system performance, then the precision and effectiveness of the scoring system is improved, but the maturation time becomes lengthy and operational impact increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing aggregate statistics (sums and counts) for each variable during a preliminary exposure phase. When new variables are introduced, their initial values are immediately available from these pre-computed statistics rather than requiring lengthy real-time exposure. This allows the scoring system to start with new variables immediately at full precision without the traditional lengthy maturation period.
2Productivity
If new variables are introduced to enhance scoring capability, then the model performance is improved, but unnecessary processing resources are consumed during maturation
Solution Approach 1:
The system performs preliminary computation of aggregate statistics during an exposure phase before the variables are actually needed for scoring. By pre-computing sums and counts for each variable and storing them, the system eliminates the need for continuous real-time processing and resource consumption during the maturation period. When new variables are introduced, their values are immediately available from pre-computed data, avoiding unnecessary processing resource usage.
3Measurement precision
If the data transaction profile is expanded with additional variables, then the model coverage and accuracy are improved, but the deployment time and cost increase
Solution Approach 1:
The patent enables rapid deployment by pre-computing and storing aggregate statistics during an exposure phase. When the profile needs to be expanded with new variables, the system can immediately initialize these variables using pre-computed statistics from the exposure phase, bypassing the traditional lengthy maturation and deployment process. This allows expanded models to be deployed rapidly without sacrificing accuracy.
Solution Approach 2:
The system creates a copy of the exposure phase data and computations, storing aggregate statistics that can be reused when new variables are introduced. Instead of re-processing raw transaction data during deployment, the system copies and utilizes pre-computed statistics, significantly reducing deployment time and computational cost while maintaining model accuracy.
Data Source
AI summary
A first set of variables is introduced into a data transaction scoring system. The first set of variables having a maturity level less than a second set of variables previously matured on the data transaction scoring system. The maturity level corresponding to an amount of exposure to the data transaction scoring system. The amount of exposure affecting a degree of precision for the data transaction scoring system. The first set of variables are introduced by expanding a data transaction profile for the data transaction scoring system including second set of variables to further include the first set of variables. Initial values are assigned for the first set of variables based on a statistical model. After such an assignment, scoring of data transactions using the expanded data transaction profile prior to maturing the data transaction system using the first set of variables can be initiated. Related apparatus, systems, techniques, and articles are also described.