ML Input Data Weighting for Volatility-Stable Model Output

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing machine learning models are affected by volatility in continuously updated data streams, leading to inaccuracies and inefficiencies in model output.

Innovation Solution

A system that identifies and weights volatile data by analyzing model output shifts, applying dampening factors to reduce the impact of older data over time, and proactively adapting to data fluctuations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If machine learning models continuously process updated data streams, then the model can adapt to new information and maintain relevance, but the model output becomes affected by data volatility leading to inaccuracies

Engineering Contradiction:

Improvemodel adaptabilityVSAvoidmodel output accuracy

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The system dynamically adjusts data weighting based on volatility detection. It continuously monitors data streams, identifies volatile periods, and adapts the influence of incoming data in real-time, allowing the model to be more responsive during stable periods and more conservative during volatile periods, thus resolving the contradiction between adaptability and reliability

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the parameter of data weighting dynamically. By detecting volatility in data streams and adjusting the weight given to new data versus historical data, the system modifies its processing parameters to maintain accuracy while adapting to new information, directly addressing the contradiction between model adaptability and output accuracy

Inventive Principle:

Principle #35Parameter changes

2Productivity

If the system applies equal weighting to all data points, then the processing is simple and efficient, but volatile data disproportionately affects model output reducing accuracy

Engineering Contradiction:

Improvedata processing efficiencyVSAvoidmodel output accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system applies different weighting qualities to different data points based on their volatility characteristics. Instead of uniform weighting, it assigns local weights that reflect the reliability of each data point, giving less weight to volatile data and more weight to stable data, thereby improving accuracy without significantly complicating the processing pipeline

Inventive Principle:

Principle #3Local quality

3Adaptability or versatility

If the system heavily weights recent data to capture current trends, then the model becomes more responsive to new information, but older valuable data is discounted too much leading to instability

Engineering Contradiction:

Improvemodel responsivenessVSAvoidmodel output stability

Core Design Contradiction:

Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The system dynamically determines weighting based on detected volatility rather than applying a fixed recency bias. During stable periods, it can incorporate more historical data for stability, while during volatile periods, it relies more on recent data for responsiveness, thus dynamically balancing adaptability and stability based on actual data conditions

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS20260073309A1System for time based monitoring and improved integrity of machine learning model input data

Publication Date: 2026.03.12 BANK OF AMERICA CORP

AI summary

Embodiments of the invention are directed to systems, methods, and computer program products for providing intelligent system and methods for identifying and weighting volatile data in machine learning data sets. The system is adaptive, in that it can be adjusted based on the needs or goals of the user utilizing it, or may intelligently and proactively adapt based on the data set or machine learning model being employed. The system may be seamlessly embedded within existing applications or programs that the user may already use to interact with one or more entities, particularly those which aid in the managing of user resources.