Media Object Effect Prediction Using Tree-Based Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for determining the effectiveness of media objects, such as commercials, are labor-intensive and costly, requiring human intervention and complex procedures, limiting their economically justifiable use to advertising effectiveness research.

Innovation Solution

A computer-implemented method using a tree-based prediction model to automatically forecast the effects of media objects by extracting signature vectors from media objects and mapping them to effect vectors, allowing for the prediction of key performance indicators (KPIs) without human intervention.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional survey methods with human intervention are used to determine media object effectiveness, then measurement precision is improved, but device complexity and loss of time increase

Engineering Contradiction:
Improveeffectiveness measurement accuracyVSAvoidsurvey procedure complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces manual survey procedures with an automated computer-implemented system that extracts signature vectors from media objects and uses machine learning models to predict effectiveness parameters. This substitution eliminates human intervention in the measurement process while maintaining prediction accuracy through trained algorithms that analyze media object properties automatically.

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

Solution Approach 2:

The system enables self-service by allowing media objects to be automatically analyzed without human involvement. The extractor automatically processes media objects to generate signature vectors, and the prediction module autonomously determines effectiveness parameters based on these vectors, making the entire process self-sufficient and eliminating the need for manual survey administration.

Inventive Principle:
Principle #25Self-service

2Measurement precision

If conventional survey methods are used to determine media object effectiveness, then measurement precision is improved, but loss of time increases

Engineering Contradiction:
Improveeffectiveness measurement accuracyVSAvoidsurvey implementation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary action by pre-training prediction models on large datasets of media objects and their effectiveness outcomes. Once trained, the models can rapidly predict effectiveness parameters for new media objects without requiring time-consuming manual surveys. The signature extraction process is also optimized to quickly generate feature vectors from media objects, enabling fast predictions.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

By replacing manual survey administration with automated computer-implemented processing, the system eliminates the time required for human respondents to complete surveys and for researchers to manually analyze results. The automated extraction and prediction process operates continuously and instantaneously, dramatically reducing the time from media object creation to effectiveness assessment.

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

3Productivity

If automated prediction methods are used to forecast media object effects, then productivity is improved, but measurement precision may worsen

Engineering Contradiction:
Improveprediction throughputVSAvoideffectiveness prediction accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The system performs extensive preliminary training of prediction models using large datasets of media objects with known effectiveness outcomes. This pre-training establishes accurate prediction capabilities before deployment, ensuring that the automated system maintains measurement precision while achieving high productivity. The models learn from historical data to accurately predict effectiveness parameters for new media objects.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system incorporates feedback mechanisms where prediction results can be compared against actual survey outcomes when available. This feedback loop allows for continuous refinement and validation of the prediction models, ensuring that automated predictions maintain high accuracy. The system can be retrained with new data to improve prediction precision over time while maintaining high throughput.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4276724A1Automatic prediction of effects of a media object
Publication Date: 2023.11.15 RTL DEUTSCHLAND GMBH
  • EP4276724A1 patent drawingFigure 1
  • EP4276724A1 patent drawingFigure 2
  • EP4276724A1 patent drawingFigure 3

AI summary

A computer-implemented method for the automatic prediction of the effects of a media object (MO) using a computer-implemented prediction module (PM), wherein the respective media object (MO) to be predicted with regard to its effects is supplied to an automatic extractor (EXT) for analysis, which automatically determines an associated signature vector (SV) from this and supplies it to the prediction module (PM), and the prediction module (PM) determines an associated effect vector (WV) on the basis of the signature vector (SV).wherein the prediction module (PM) uses a tree-based prediction model to determine the effect vector (WV) containing the respective parameters for the effectiveness of the respective media object (MO) from the signature vector (SV) containing the properties of the respective media object (MO), and was previously trained by means of a computer-implemented method for training the computer-implemented prediction module (PM) for the automatic prediction of effects of a media object (MO) according to the present invention using a database of available media objects (DB-MO) to which an effect vector (BMV) known to be correct with regard to the effect of the respective media object is assigned, and the aforementioned training method.