Automated Forecasting Model Selection via Econometric Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current forecasting models face challenges in selecting the most accurate variables due to numerous influencing factors and insufficient data, leading to inaccurate predictions and wastage of resources.

Innovation Solution

A system utilizing a metadata layer and data layer to enhance query processing and 'what-if' scenario analysis, allowing for faster multi-dimensional querying and variable selection through a web-based GUI, which includes econometric modeling, data diagnostics, and visualization, enabling the development of accurate marketing plans.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If multiple variables are included in the forecasting model to account for all influencing factors, then the model's comprehensiveness is improved, but the model's accuracy deteriorates due to difficulty in selecting the most accurate variables

Engineering Contradiction:
Improvemodel comprehensivenessVSAvoidforecasting accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent replaces manual variable selection and model building processes with automated econometric modeling and machine learning algorithms. The system automatically processes historical data, identifies relevant variables, and generates forecasting models without requiring manual intervention, thereby resolving the contradiction between model comprehensiveness and accuracy by using computational methods to objectively select the most accurate variables from multiple influencing factors

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent creates virtual copies of historical data and scenarios to test and validate forecasting models. By using historical data as a basis for creating test scenarios and comparing model predictions against actual outcomes, the system can evaluate and select the most accurate variables and model configurations, thereby improving forecasting accuracy while maintaining comprehensive variable consideration

Inventive Principle:
Principle #26Copying

2Measurement precision

If comprehensive data analysis is performed to select the most accurate variables, then the forecasting accuracy is improved, but the processing time increases

Engineering Contradiction:
Improveforecasting accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary processing of historical data by organizing it into standardized formats and pre-processing it before actual forecasting is needed. The system pre-loads and structures historical data in advance, so when forecasting is required, the data is already prepared and ready for rapid analysis, thereby reducing the time required for comprehensive data analysis while maintaining forecasting accuracy

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses automated computational algorithms and machine learning models to perform comprehensive data analysis rapidly without manual intervention. The system employs efficient algorithms that can process large volumes of historical data and identify accurate variables much faster than manual analysis, thereby resolving the contradiction between comprehensive analysis and processing time through automated computational methods

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP2273431B1Model determination system
Publication Date: 2023.01.25 ACCENTURE GLOBAL SERVICES LTD
  • EP2273431B1 patent drawingFigure 1
  • EP2273431B1 patent drawingFigure 2
  • EP2273431B1 patent drawingFigure 3

AI summary

A system includes a variable determination module determining a variable operable to be used for the final model and also determining a modification to the at least one variable. An assumption determination module determines an assumption operable to be used for the final model. The assumption includes a transformation for the variable describing how the variable impacts the objective or how the variable impacts another variable operable to be used in the final model. The assumption module also determines a modification to the assumption. A model generator generates a candidate model using the variable and the assumption, and generates a new candidate model using the modified assumption, the new variable or the modification to the variable. The candidate model or the new candidate model is operable to be selected as the final model based on at least one of a statistical measure and an indication of relevance for the variable in the candidate model and the new candidate model.