ML Model Generation With Correlated Datasets and Hyperparameter Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The generation, training, and testing of machine learning (ML) computer models, particularly in weather-based prediction, is a labor, time, and resource-intensive process due to variability and unpredictability, requiring extensive manual experimentation and often failing to exhaustively explore all possibilities for achieving accurate AI solutions.
Innovation Solution
An automated computing tool and process that groups datasets based on correlation, generates ML computer model instances with varied threshold and hyperparameter combinations, trains, tests, and evaluates them, and selects the best performing models, addressing sparsity and variability in input datasets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual experimentation is used to generate, train, and test ML computer models, then model accuracy can be improved through extensive experimentation, but the process becomes labor-intensive, time-consuming, and resource-intensive
Solution Approach 1:
The system performs self-service by automatically generating, training, and evaluating multiple ML model instances with different configurations without human intervention. The automated workflow includes generating dataset groups, creating model instances with varying thresholds and hyperparameters, training them, and selecting optimal models based on performance metrics, thereby eliminating manual experimentation while maintaining accuracy.
Solution Approach 2:
The system applies parameter changes by systematically varying thresholds and hyperparameters across multiple ML model instances. Each model instance is configured with different parameter combinations, allowing the system to explore the parameter space automatically and identify optimal configurations without manual intervention, thus improving accuracy while reducing time consumption.
2Reliability
If extensive manual experimentation is conducted to generate and test multiple ML model configurations, then optimal model selection can be improved, but the resource requirements and complexity increase significantly
Solution Approach 1:
The system applies segmentation by dividing the model development process into distinct automated stages: dataset group generation, model instance creation with specific parameter configurations, training execution, performance evaluation, and model selection. This structured segmentation manages complexity by automating each stage while maintaining reliable model selection through systematic exploration of configurations.
Solution Approach 2:
The system uses copying by generating multiple copies of ML model instances with different parameter configurations automatically. Instead of manually creating and testing each configuration, the system creates numerous model instance copies through automated workflows, trains them in parallel or sequence, and selects the best performing ones, thereby improving selection quality while managing complexity through automation.
3Manufacturing precision
If multiple ML model instances with different thresholds andhyperparameters are generated and evaluated, then the quality of model deployment can be improved, but the computational resources and time required increase
Solution Approach 1:
The system applies partial or excessive action by generating and evaluating multiple ML model instances with different configurations, then selecting only the top-performing models for deployment. Rather than exhaustively testing every possible configuration, the system performs sufficient experimentation to identify optimal models and stops there, achieving high configuration precision while managing computational resource consumption through selective evaluation.
Data Source
AI summary
Mechanisms are provided to automatically generate a machine learning (ML) computer model. The mechanisms automatically generate a plurality of aggregated dataset groups, each having original dataset(s) grouped together based on a degree of correlation between characteristics of each of the original datasets. The mechanisms automatically generate, for each aggregated dataset group, a plurality of ML computer model instances, each being a ML computer model configured with a different combination of thresholds and hyperparameters than other ML computer model instances. The plurality of ML computer model instances are executed to generate performance metric information for each ML computer model instance. The performance metric information is analyzed to select a set of ML computer model instances for the aggregated dataset. The mechanisms select one or more ML computer model instances from across all of the sets of ML computer model instances as a candidate for deployment to a decision support computing system.