Resource Requirement Estimation Using ML Clustering and Simulation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods struggle to accurately estimate resource requirements, especially in dynamic environments where unforeseen circumstances occur, as machine learning predictions for long-term scenarios are often inaccurate and fail to account for uncertainty.

Innovation Solution

A method involving a preparation phase with machine learning training and clustering, where historical data is used to train a model and determine feature subsets, and a simulation phase that uses cluster distributions and adjustment parameters to estimate future resource needs, reducing complexity and improving accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If machine learning methods are used for long-term resource requirement predictions, then automation is improved, but prediction accuracy deteriorates

Engineering Contradiction:
Improveautomation of resource requirement estimationVSAvoidprediction accuracy
Core Design Contradiction:
Extent of automationVSMeasurement precision

Solution Approach 1:

The patent segments the resource requirement estimation process into two distinct phases: a preparation phase that trains a machine learning model on historical data, and a simulation phase that uses clustering and scenario analysis for future predictions. This segmentation allows the system to leverage automation in the preparation phase while using more controlled, accurate methods in the simulation phase, thereby resolving the contradiction between automation and prediction accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies preliminary action by conducting extensive machine learning model training and data preprocessing in the preparation phase before the actual prediction is needed. This preliminary work includes training the model on historical data and preparing clustering structures, so that when long-term predictions are required, the system can rely on pre-computed, accurate scenarios rather than performing complex calculations in real-time, thus maintaining both automation and accuracy.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If machine learning models are trained on all historical data, then measurement precision is improved, but device complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts and utilizes only the most relevant features from historical data through the clustering process. By identifying key patterns and correlations in the data during the preparation phase, the system extracts essential information needed for accurate predictions while discarding redundant data. This extraction approach maintains prediction accuracy while significantly reducing the computational complexity required for long-term simulations.

Inventive Principle:
Principle #2Taking out (Extraction)

3Device complexity

If steady-state system approaches are used, then device complexity is reduced, but adaptability deteriorates

Engineering Contradiction:
Improvesystem simplicityVSAvoidresponse to unforeseen circumstances
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

The patent introduces dynamics by implementing a simulation phase that can adapt to different scenarios and unforeseen circumstances. While the preparation phase uses a relatively simple, steady-state machine learning model trained on historical data, the simulation phase dynamically adjusts predictions by considering multiple possible future scenarios, correlations between features, and potential disruptions. This dynamic approach maintains system simplicity while significantly improving adaptability to changing conditions.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4411605A1Controlling resource requirements
Publication Date: 2024.08.07 AMADEUS SAS
  • EP4411605A1 patent drawingFigure 1~2
  • EP4411605A1 patent drawingFigure 3~5
  • EP4411605A1 patent drawingFigure 6~7

AI summary

A computerized method of estimating resource requirements in an environment is presented. The method comprises a preparation phase and a simulation phase, wherein the preparation phase comprises a machine learning training phase and a clustering phase. The machine learning training phase trains a machine learning model to predict a resource requirement. Thereby, a subset of features is extracted. The clustering phase determines clusters in the subset of features, a correlation coefficient and least one identifying parameter of a distribution of the feature values of the subset of features. Finally, the simulation phase determines a distribution for a feature, selects at least one value for a feature and uses the second machine learning model to estimate a resource requirement in the environment in at least one time period the future.