Ensemble Prediction Model Pool Using Post-Modulation Outputs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ensemble prediction methods using supervised machine learning operations incur high costs due to the generation of multiple ensemble models.
Innovation Solution
A prediction device and method that manages a model pool including a first model and one or more second models generating post-modulation outputs, allowing prediction processing using a plurality of models without additional learning, thereby reducing costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple ensemble models are generated using supervised machine learning operations, then prediction accuracy is improved, but cost increases
Solution Approach 1:
The patent segments the model generation process into two distinct parts: (1) a base model generated through supervised machine learning, and (2) multiple post-modulation models generated by applying different modulation functions to the base model's output. This segmentation allows the expensive supervised learning process to be performed only once, while the multiple ensemble models are created through cheaper post-processing operations, thereby resolving the contradiction between prediction accuracy and cost.
Solution Approach 2:
The patent performs preliminary action by generating the base model through supervised machine learning in advance, before creating the ensemble models. The base model serves as a pre-trained foundation that can be modulated multiple times to create diverse ensemble models without requiring additional supervised learning processes. This preliminary action reduces the overall cost while maintaining prediction accuracy.
2Adaptability or versatility
If multiple ensemble models are generated using supervised machine learning operations, then model diversity is improved, but complexity increases
Solution Approach 1:
The patent segments model diversity generation into a single supervised learning phase for the base model, followed by multiple modulation phases using different functions (e.g., linear transformations, non-linear functions, threshold operations). This segmentation achieves model diversity through simple post-processing operations rather than through complex additional training processes, thereby reducing system complexity while maintaining adaptability.
Solution Approach 2:
The patent achieves model diversity by changing parameters and transformation functions applied to the base model output rather than by creating fundamentally different model architectures. Different modulation functions (e.g., varying coefficients, different mathematical transformations) are applied to the same base model to generate diverse ensemble models, which simplifies the overall system compared to training multiple independent models from scratch.
Data Source
AI summary
A prediction device includes at least one memory storing instructions, and at least one processor configured to execute the instructions to manage a model pool including a first model and one or more second models that generate one or more post-modulation outputs with reference to an output of the first model, and execute prediction processing using a plurality of models included in the model pool for decision making.