Agent Attribute Blending for Diverse Simulation Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing agent-generation systems fail to create a diverse set of agents that accurately reflect real-world conditions, particularly in purchasing behavior simulations, due to overly strong correlations between attributes such as age and income.
Innovation Solution
A method for assigning attributes to agents in agent-based models involves blending random numerical values with initial values, using equations like g(u1, u2) = {(u1+u2)/2 if u1+u2 < 1, (2-(2-u1-u2))/2 otherwise, to create a uniformly distributed variable that maintains correlation while introducing variability, allowing for more realistic and diverse agent characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If strong correlations are assigned between agent attributes (e.g., age and income), then the model structure is simple and easy to implement, but the agents lack diversity and do not accurately reflect real-world conditions
Solution Approach 1:
The patent applies parameter changes by transforming the correlation strength between agent attributes from strong to weak. Specifically, it uses blending functions to create intermediate values that combine the first attribute value with random values, resulting in a correlation parameter that is weaker than the original but stronger than complete randomness. This resolves the contradiction by adjusting the correlation parameter to achieve both implementability and realistic diversity.
Solution Approach 2:
The patent introduces an intermediary blending function that mediates between the first attribute value and random values. This intermediary mechanism (the blending function g(u1, u2)) creates a transition zone that maintains some correlation structure while introducing sufficient variability. The intermediary allows the system to balance model simplicity with agent diversity by controlling the degree of blending.
2Adaptability or versatility
If random values are completely mixed with initial values, then agent diversity is maximized, but the correlation structure between attributes is lost
Solution Approach 1:
The patent changes the correlation parameter from complete randomness to a controlled weak correlation. The blending function creates a correlation strength that is neither zero nor maximum, but somewhere in between. This allows the agent attributes to maintain some structural correlation (stability) while introducing sufficient diversity through the random component.
Solution Approach 2:
The patent applies partial action by using a blending function that incorporates both the first value and random values, rather than completely replacing one with the other. The blending parameter controls the degree of mixing, ensuring that some correlation structure is preserved while sufficient diversity is introduced. This partial mixing approach resolves the contradiction between maintaining structure and achieving diversity.
Data Source
AI summary
A method for assigning an attribute to an agent for use in an agent-based model, including receiving or selecting a first numerical value for assigning an attribute of an agent in an agent-based model. The method further includes receiving or selecting a random numerical value, blending the first and random numerical values, resulting in an intermediate value, and blending the intermediate result with the first numerical value, or the random numerical value, or with a second numerical value, resulting in an end value. The method further includes utilizing the end value to assign the attribute of the agent in the agent-based model.