Causality Score Calibration for Debt Collection Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for scoring the effects of operations on users are limited by high bias and low variance, leading to inaccurate predictions of future scores, particularly in debt collection scenarios where increased operations result in higher costs without necessarily improving debt collection rates.
Innovation Solution
An information processing device that employs a calibration function to combine causality scores from both high-bias, low-variance and low-bias, high-variance models, adjusting the first causality score to produce a more accurate third causality score, thereby optimizing operation effectiveness and cost management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a machine learning model is trained with strong dependence on past user action tendencies and results, then the model can capture historical patterns, but the model's ability to accurately predict future scores deteriorates due to high bias and low variance
Solution Approach 1:
The patent divides the single machine learning model into two separate models: a first model that captures historical patterns with high bias and low variance, and a second model that provides diverse predictions with low bias and high variance. This segmentation allows each model to specialize in different aspects of prediction, resolving the contradiction between pattern capture and prediction accuracy.
Solution Approach 2:
The patent creates a composite scoring system by combining outputs from two different machine learning models through a calibration function. This composite approach integrates the strengths of both models - the historical pattern recognition of the first model and the predictive flexibility of the second model - to achieve superior overall prediction accuracy.
2Productivity
If operations are increased to improve debt collection rates, then more users can be reached, but operational costs increase without necessarily improving collection rates
Solution Approach 1:
The patent applies local quality by providing individualized causality scores for each user based on their specific attributes and historical behavior. This allows operations to be tailored to each user's likelihood of responding, concentrating resources on users with high predicted collection probability rather than applying uniform operations to all users.
Solution Approach 2:
The patent changes the parameter of operation targeting by using calibrated causality scores to identify and prioritize users with high predicted response likelihood. This parameter change from random or uniform targeting to score-based selective targeting improves collection rates while reducing operational costs by focusing resources efficiently.
Data Source
AI summary
An information processing device includes a first effect estimating unit for obtaining a first causality score indicating an effect that a predetermined operation has on a user, by inputting an attribute of the user to a first model, a second effect estimating unit for obtaining a second causality score indicating the effect, by inputting the attribute of the user to a second model, a calibration function deciding unit for deciding a calibration function for calibration of the first causality score, based on the first causality score and the second causality score calculated for each of a plurality of users, and a third effect estimating unit for deciding a third causality score indicating the effect on an object user, by applying the first causality score, which is calculated with regard to the object user, to the calibration function.