Accounts Receivable Model Selection Using Grace-Period MCC
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models struggle to accurately predict accounts receivable payment delays due to wide variations among customers, class imbalance, and increased variability, leading to unpredictable outcomes and inability to distinguish between delayed and non-delayed payments.
Innovation Solution
The solution involves segmenting customers based on the median-based coefficient of variation (MCoV) to create tailored models for each group, incorporating grace periods, and using the Matthews' Correlation Coefficient (MCC) to select the most accurate model for predicting payment delays, while generating a customer reliability score.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single machine learning model is used to predict AR payments for all customers, then the model structure is simple, but the prediction accuracy deteriorates due to wide variations among customers and class imbalance
Solution Approach 1:
The patent segments customers into distinct groups based on their payment behavior characteristics and divides the prediction task into multiple specialized models, each tailored to a specific customer segment. This segmentation allows each model to focus on the unique patterns of its target group, thereby improving overall prediction accuracy while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The patent applies local quality by creating customized prediction models for different customer segments, where each model is optimized for the specific characteristics and payment patterns of its target group. This ensures that each customer segment receives predictions tailored to their local needs and behaviors, improving accuracy without requiring a completely different approach for each group.
2Measurement precision
If multiple models with different grace periods are created for different customer segments, then the prediction accuracy improves, but the model selection and management complexity increases
Solution Approach 1:
The patent employs parameter changes by systematically varying the grace period parameter across different models to optimize performance for different customer segments. By adjusting this key parameter and other model-specific parameters, the system creates a family of models that can be selectively deployed based on segment characteristics, improving accuracy while managing complexity through parameterized model generation.
Solution Approach 2:
The patent implements feedback mechanisms that continuously monitor model performance on validation data and use this information to guide model selection and refinement. This feedback loop allows the system to automatically identify which models perform best for each customer segment and adjust future model training accordingly, reducing manual selection complexity while maintaining high accuracy.
3Reliability
If the model is trained to distinguish between defaulting and non-defaulting customers, then the predictive capability improves, but the model becomes more complex due to class imbalance and disparate variability in payment delays
Solution Approach 1:
The patent segments customers based on their payment delay patterns and class distribution, creating separate models for different segments. This segmentation simplifies the training task by focusing on homogeneous groups with similar characteristics, making it easier to distinguish between defaulting and non-defaulting customers within each segment while reducing overall model complexity through modular design.
Solution Approach 2:
The patent uses parameter changes including adjusted class weights, modified loss functions, and varied grace period parameters to handle class imbalance and disparate variability in payment delays. These parameter adjustments enable the models to effectively distinguish between defaulting and non-defaulting customers without requiring overly complex architectures, as the parameter tuning compensates for the imbalance and variability issues.
Data Source
AI summary
Embodiments predict a target variable for accounts receivable using a machine learning model. For a first customer, embodiments receive a plurality of trained ML models corresponding to the target variable, the plurality of trained ML models trained using the historical data and comprising a first trained model having no grace period for the target variable and two or more grace period trained models, each grace period trained model having different grace periods for the target variable. Embodiments determine a Matthews' Correlation Coefficient (“MCC”) for the first trained model. When the MCC for the first trained model is low, embodiments determine the MCC for each of the grace period trained models, and when one or more MCCs for each of the grace period trained models is higher than the MCC for the first trained model, embodiments select the corresponding grace period trained model having a highest MCC.