Hierarchical ESG Index Modeling for Supplier Risk Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current index modeling systems for supply chains face challenges in accurately predicting ESG index scores due to subjective weight assignment in sub-indicators, leading to complex indicator values and unclear feature contributions, which complicates the determination of true indicator effects and risks.
Innovation Solution
The proposed index modeling system employs a bottom-to-top approach with a hierarchical structure, using multiple linear regression models and feature importance calculations to predict attribute values for ESG indexes, selecting only indicators that clear correlation thresholds, and providing explainable results by analyzing sub-indicators and indicators individually.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple linear regression models with feature importance calculations are used to predict ESG index scores, then measurement precision and explainability are improved, but device complexity increases due to hierarchical structure requirements
Solution Approach 1:
The model hierarchically segments ESG index prediction into multiple levels: first predicting country-level ESG scores, then supplier-level scores, and finally commodity-level aggregated scores. Each level uses separate regression models with progressively refined features, breaking down the complex prediction task into manageable segments that improve precision while maintaining interpretability
Solution Approach 2:
The patent introduces a hierarchical dimension to the prediction model, adding country, supplier, and commodity levels beyond traditional flat structures. This dimensional expansion allows the model to capture relationships at multiple scales simultaneously, improving measurement precision by accounting for contextual factors at each hierarchical level
2Manufacturing precision
If only indicators clearing correlation thresholds are selected for modeling, then manufacturing precision of the model is improved, but loss of information occurs by excluding sub-indicators
Solution Approach 1:
The patent extracts and separately models high-correlation indicators that clear threshold requirements, isolating them for focused regression analysis. By extracting only the most relevant indicators while systematically excluding lower-correlation ones, the model achieves higher construction precision without arbitrarily discarding data, as the exclusion process itself is informative
Solution Approach 2:
The patent changes the correlation threshold parameter dynamically at different hierarchical levels and for different indicator types. By adjusting this parameter based on data characteristics and model performance requirements, the system optimizes the balance between including sufficient information and maintaining model precision, transforming a static filtering rule into an adaptive selection mechanism
3Ease of operation
If subjective weight assignment is used in sub-indicators, then ease of operation is improved, but measurement precision deteriorates due to unclear feature contributions
Solution Approach 1:
The regression models automatically determine feature importances and weights based on statistical relationships in the data, eliminating the need for manual subjective weight assignment. The model serves itself by identifying which indicators and sub-indicators have the strongest predictive power for ESG scores, with feature importance metrics providing objective clarity on contributions
Solution Approach 2:
The patent implements feedback loops where model predictions are validated against actual ESG data, and feature importances are recalculated based on prediction performance. This continuous feedback refines the objective weight assignments, improving measurement precision by adjusting weights based on their actual predictive value rather than subjective judgment
Data Source
AI summary
An index modeling system that generates index models that predict values of an attribute of a supply chain for a commodity is disclosed. The index models are generated from indicator data that includes data related to multiple indicators and a plurality of sub-indicators of the index arranged in a hierarchical structure. Accordingly, the index values can be predicted for different entities at different levels in the hierarchical structure. The predicted index values can be used to automatically generate a filtered list of suppliers who can be used for procurement based on comparisons of the predicted attribute values of the suppliers with a predetermined attribute threshold value.