Asset Valuation Visualizations for Interpretable Machine Learning Outputs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industry professionals face difficulty in understanding the complex processes behind predicted asset valuations, making it challenging to take informed actions based on these valuations.
Innovation Solution
A system and method for creating a valuation model that generates explanations and visualizations of the factors influencing the output, enabling professionals to understand how valuations are derived and presented in a digestible manner.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If increasingly complicated tools are used to determine predicted asset values, then the precision of valuation is improved, but the ease of operation deteriorates as it becomes difficult for professionals to understand what is causing the predicted value
Solution Approach 1:
The patent introduces visualization tools as an intermediary between the complex machine learning valuation model and the user. These visualizations translate complex model outputs into intuitive graphical representations that show which factors drive predicted values, making the black-box model interpretable without simplifying the underlying complex algorithms
Solution Approach 2:
The patent replaces traditional mechanical explanation methods (textual descriptions, numerical tables) with visual representation systems. By substituting conventional explanation mechanisms with graphical visualizations, the system maintains high valuation precision while dramatically improving user comprehension of model outputs
2Reliability
If the complexity of valuation tools increases, then the reliability of predicted values is improved, but the ease of operation worsens making it difficult for industry professionals to understand the valuation process
Solution Approach 1:
Visualization tools serve as an intermediary layer that bridges the gap between reliable but complex machine learning models and users who need to understand and trust the outputs. The visualizations do not compromise model reliability while making the decision-making process transparent and interpretable
Solution Approach 2:
The patent employs color-coded visualizations to represent different factors influencing asset valuations. By using color variations to encode information about factor importance and direction of influence, the system enables quick visual comprehension of complex valuation drivers without reducing model reliability
Data Source
AI summary
A method for providing a visualization explaining an output includes generating, by an engine and using a trained model, a predicted net operating income (NOI) and an explanation dataset for each asset in an asset dataset that is used as an input to the trained model and the explanation dataset includes deviations from a baseline dataset. The method also includes generating, by the engine, for at least one of the assets in the combined dataset, the visualization based on the predicted NOI and the explanation dataset. The visualization includes a total net impact based on a total difference between the predicted NOI and the baseline dataset and the deviations. Each of the deviations is based on one type from an economic and demographic dataset (EDD), and the deviations are displayed in an order based on the magnitude of the deviation.