Time-Segmented ML Models for Time-Varying Prediction Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning models fail to account for time-restricted variability in training data, leading to inaccurate predictions or decisions regarding time-varying data.
Innovation Solution
The development of time-restricted machine learning models that segment training data into multiple time periods, training a model for each period to identify attributes correlated with the likelihood of a target action, and selecting attributes and time periods to perform operations that reduce the likelihood of the target action.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional machine learning models are used without time segmentation, then the model structure is simple and training is fast, but the prediction accuracy for time-varying data deteriorates because the models cannot account for time-restricted variability in features
Solution Approach 1:
The patent segments the training data into multiple time periods (e.g., day parts, hours, or custom time windows) and trains separate machine learning models for each time segment. This allows the system to capture time-varying feature relevance while maintaining manageable model complexity through modular architecture.
Solution Approach 2:
The patent implements dynamic model selection where the system automatically selects which time-specific model to apply based on the current time context. This dynamic approach allows the system to adapt to changing feature relevance over time without requiring a single complex static model.
2Measurement precision
If time-restricted machine learning models are implemented with multiple time periods, then the prediction accuracy for time-varying data improves, but the training time and computational resources increase
Solution Approach 1:
By dividing the training process into separate time segments, the system can train multiple simpler models in parallel rather than training one large complex model sequentially. This segmentation reduces the computational burden per model while maintaining overall accuracy.
Solution Approach 2:
The patent adjusts training parameters such as sample size, feature selection, and model hyperparameters for each time period based on the specific characteristics of that time segment. This optimization reduces training time for each individual model while preserving prediction accuracy.
3Adaptability or versatility
If time-specific models are trained for each period, then the ability to identify event-dependent features improves, but the device complexity and data processing requirements worsen
Solution Approach 1:
The patent segments both the data and the modeling process by time period, allowing feature relevance to be determined independently for each time segment. This segmentation enables the system to identify event-dependent features specific to different times without requiring a single complex feature set.
Solution Approach 2:
The patent employs a universal framework and common data processing pipeline that handles multiple time periods consistently. This multi-functional approach allows the same system architecture to serve different time-specific modeling needs, reducing overall system complexity despite handling multiple time periods.
Data Source
AI summary
A computer system receives a training data set that includes representations of a population of entities that are associated with a target action. Each representation in the training data set including a set of attributes of a respective entity and a time the target action associated with the respective entity occurred. The computer system segments the training data set into multiple time periods, then trains a model to identify, for each of the multiple time periods, an attribute that is correlated with a likelihood that the target action will occur with respect to an entity having the identified attribute during a corresponding time period. The computer system uses the model to perform a first operation for a first set of entities in a first time period and perform a second operation for a second set of entities in a second time period.