Parallelized Ensemble Model Scoring via Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Ensemble models, such as those comprising thousands of decision trees, often take a long time to produce predictions, which is undesirable for real-time applications like credit card fraud detection where timely results are critical.
Innovation Solution
The method involves extracting features from base models in an ensemble model, clustering them based on time performance, and assigning these clusters to parallel computation units to optimize prediction processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ensemble models include thousands of base models to improve accuracy, then prediction accuracy is improved, but prediction time increases significantly
Solution Approach 1:
The patent divides the large ensemble model into multiple smaller clusters of base models. Each cluster is processed independently by separate computation units, allowing parallel execution. This segmentation reduces the prediction time while maintaining the collective accuracy benefit of having thousands of base models.
Solution Approach 2:
The patent introduces a new dimension of parallel processing by distributing base models across multiple computation units that operate simultaneously. Instead of sequential processing within a single unit, the system utilizes spatial distribution and concurrent execution to reduce overall prediction time.
2Measurement precision
If ensemble models are processed sequentially to ensure accurate results, then prediction accuracy is maintained, but processing speed decreases
Solution Approach 1:
The patent segments the ensemble model processing into multiple independent clusters that can be executed in parallel. Each computation unit processes its assigned cluster independently, ensuring accurate results while achieving high processing speed through concurrent execution of multiple segments.
Solution Approach 2:
The patent merges the results from multiple parallel computation units to produce the final prediction. By combining the outputs of independently processed clusters, the system maintains the accuracy benefits of comprehensive ensemble evaluation while achieving high processing speed through parallel execution.
3Productivity
If base models are distributed evenly across computation units to balance workload, then system efficiency is improved, but load balancing becomes complex
Solution Approach 1:
The patent applies local quality by creating clusters with homogeneous characteristics within each computation unit. Base models with similar properties are grouped together, ensuring that each computation unit handles workloads of comparable complexity. This local homogeneity simplifies load balancing while maintaining high system efficiency.
Solution Approach 2:
The patent changes the organizational parameter from individual model distribution to cluster-based distribution. By grouping base models into clusters and distributing these clusters across computation units, the system achieves better load balancing with reduced complexity compared to managing individual model assignments.
Data Source
AI summary
Provided are a computer-implemented method, a system, and a computer program product. The method comprises extracting features from a plurality of base models in an ensemble model. The plurality of base models are configured to provide respective prediction results. The ensemble model is configured to provide an overall prediction result from the prediction results of the plurality of base models. The features are associated with time performance of the base models. The method further comprises clustering the plurality of base models into a plurality of clusters based on the extracted features. The method further comprises assigning the plurality of base models to a plurality of parallel computation units based on the plurality of clusters.