Economic Network Flow Estimation via Iterative Probabilistic Inversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for analyzing economic networks are limited by the lack of complete data on relationships between entities, as many economic activities are private and only partial information is publicly available, hindering the ability to perform advanced network analytics that require complete characterization of internal network values.
Innovation Solution
A method is developed to estimate flows between entities in an economic network using a combination of public and private data sources, employing a simulation method or closed-form solution to iteratively update the network map, incorporating placeholder entities and using probabilistic inversion procedures like iterative proportional fitting to ensure maximum likelihood estimation, even with limited data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complete data on all economic relationships is collected, then network characterization accuracy is improved, but data availability and accessibility deteriorate
Solution Approach 1:
The patent introduces an intermediary estimation model that mediates between the incomplete data and the requirement for complete network characterization. The model uses observable proxy variables and statistical relationships to infer unobserved economic flows, effectively acting as a bridge that allows accurate network analysis without requiring complete direct data on all relationships.
Solution Approach 2:
The patent replaces the mechanical approach of directly collecting and processing complete relationship data with a statistical inference mechanism. Instead of mechanically gathering all economic transaction data, the system uses probabilistic models and observed patterns to substitute for missing data, achieving network characterization through statistical estimation rather than complete data collection.
2Ease of manufacture
If static binary relationship information is used, then data collection simplicity is improved, but network analytics capability deteriorates
Solution Approach 1:
The patent transforms static binary relationship data into dynamic flow estimates by introducing time-varying characteristics and probabilistic transitions. The model captures the dynamic nature of economic relationships through transition probabilities and flow rates, enabling advanced analytics while maintaining relatively simple data collection requirements.
Solution Approach 2:
The patent changes the parameters of relationship representation from simple binary existence to multi-dimensional flow characteristics. By introducing parameters such as flow magnitude, transition probabilities, and temporal dynamics, the system enhances network analytics capability without fundamentally changing the simplicity of data collection.
3Measurement precision
If iterative estimation methods are applied, then estimation accuracy is improved, but computational complexity and time required deteriorate
Solution Approach 1:
The patent implements feedback mechanisms in the iterative estimation process, where results from each iteration are fed back to refine subsequent estimates. This feedback loop enables the system to converge on accurate estimates efficiently by learning from previous iterations and adjusting the estimation trajectory, reducing overall computational time while maintaining high accuracy.
Solution Approach 2:
The patent dynamically adjusts estimation parameters and convergence criteria based on the iterative process. By monitoring estimation progress and adapting parameters such as convergence thresholds and iteration steps, the system achieves accurate results with optimized computational time, avoiding unnecessary iterations while ensuring sufficient precision.
Data Source
AI summary
Obtaining data including company data and their known connections through public sources will necessarily result in an incomplete matrix of customer-supplier relationships containing some mixture of known relationships with known values, known relationships with unknown values, and unknown relationships with unknown values. A method and system is presented to obtain a best estimate of all unknown values given the known information in the network, including an amount that is assigned to unknown entities to be discovered later.