Incremental Electronic Currency Impact Prediction Using Counterfactual Models

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current methods for determining the value contributed by a merchant data object for resource allocation in electronic promotion and marketing systems are inefficient, leading to sub-optimal utilization of resources due to inaccurate predictive incremental electronic currency impacts.

Innovation Solution

The use of specially configured machine learning models, including counterfactual and predictive models, to generate accurate predictive incremental electronic currency impacts for merchant data objects, enabling precise resource allocation and ranking.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional resource allocation methods are used for merchant data objects, then system simplicity is maintained, but resource allocation accuracy deteriorates leading to sub-optimal utilization

Engineering Contradiction:

Improvepredictive incremental electronic currency impact accuracyVSAvoidmachine learning model system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the resource allocation problem into multiple independent machine learning models: a counterfactual model for baseline prediction, a predictive model for future value estimation, and an ensemble model for final ranking. Each model handles a specific aspect of the prediction task, improving overall accuracy while maintaining manageable complexity through modular design.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces intermediary computational components including feature engineering modules that preprocess merchant data, hyperparameter optimization layers that tune model performance, and ensemble aggregation mechanisms that combine predictions. These intermediaries bridge the gap between raw data and final resource allocation decisions, enhancing measurement precision.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Loss of energy

If computing resources are allocated without accurate predictive models, then resource allocation speed is maintained, but resource waste increases due to poor assignment

Engineering Contradiction:

Improvecomputing resource wasteVSAvoidmodel training and prediction time

Core Design Contradiction:

Loss of energyVSLoss of time

Solution Approach 1:

The system performs preliminary actions by pre-training the counterfactual and predictive models on historical merchant data before actual resource allocation decisions are needed. Feature engineering and hyperparameter optimization are conducted in advance, so that when resource allocation is required, the models are already prepared and can provide rapid predictions, minimizing time loss while preventing resource waste through accurate forecasting.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent dynamically adjusts model parameters including confidence thresholds for resource allocation, ensemble weighting factors, and feature selection criteria based on data availability and computational constraints. This allows the system to optimize the balance between prediction accuracy and computation time, reducing both resource waste and time loss adaptively.

Inventive Principle:

Principle #35Parameter changes

3Reliability

If multiple machine learning models are implemented for accurate prediction, then predictive accuracy improves, but computational complexity increases

Engineering Contradiction:

Improveresource allocation reliabilityVSAvoidmodel ensemble system complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The ensemble system is segmented into specialized models with distinct functions: the counterfactual model handles causal inference, the predictive model forecasts future performance, and the ranking model synthesizes predictions. This segmentation improves reliability by ensuring each aspect is handled by expertise-tuned models while keeping individual model complexity manageable.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The machine learning system is designed with universal components that serve multiple functions: the same feature engineering pipeline supports all models, the ensemble framework can accommodate different model types, and the deployment architecture handles both training and inference uniformly. This multi-functionality reduces overall system complexity despite using multiple specialized models.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS12118496B1Merchant incremental electronic impact value prediction and ranking using multiple machine learning models

Publication Date: 2024.10.15 BYTEDANCE INC

AI summary

Methods, apparatus, systems, and computer program products are disclosed for utilizing specially configured machine learning models to generate incremental currency value(s) associated with one or more target merchant data objects. Some embodiments, based on one or more market record sets, identify an actual electronic currency value for a total merchant data object set, and include a counterfactual model configured to generate a counterfactual electronic currency value for use in determining a counterfactual incremental electronic currency impact, and in some embodiments for ranking other models. Embodiments, additionally or alternatively, include an incrementality-trained ensemble model for generating a predictive incremental electronic currency impact. The incrementality-trained ensemble model may be trained to predict based on the rankings of the outputs of the counterfactual model. Embodiments may further rank target merchant data objects and perform one or more additional actions, including assigning the target merchant data objects to sales account data structures for management.