Feature Calculation Loop for Change-Driven Recalculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Feature calculation in analytical processing is computationally expensive and time-consuming due to the size and complexity of feature computations, which affects both training data and analytical model inputs, preventing real-time applications and leading to inefficient recalculations of unchanged data.
Innovation Solution
Implementing a feature calculation loop that tracks data changes and recalculates only necessary features, using listeners to maintain a feature-to-attribute map and apply a significance threshold, enabling continuous and efficient feature generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If feature calculation is performed on all data, then completeness of feature computation is improved, but computational cost and time consumption increase
Solution Approach 1:
The patent segments the feature calculation process by dividing features into different groups based on their volatility characteristics. Only volatile features that have actually changed are recalculated, while non-volatile features are skipped. This segmentation resolves the contradiction by maintaining completeness for necessary features while avoiding redundant calculations for unchanged features.
Solution Approach 2:
The patent introduces a volatility parameter to characterize feature properties. By changing the calculation approach based on this parameter (recalculating volatile features, skipping non-volatile ones), the system achieves both completeness where needed and efficiency where possible, resolving the time-cost contradiction.
2Speed
If feature calculation is performed frequently, then real-time analytics capability is improved, but computational resource consumption increases
Solution Approach 1:
The patent implements a dynamic feature calculation approach where the calculation frequency and scope adjust based on data change detection. The system continuously monitors for changes and only triggers calculations when necessary, making the processing dynamic rather than static. This resolves the contradiction by enabling real-time responsiveness only when data changes occur.
Solution Approach 2:
The patent employs a feedback mechanism where the system monitors data changes and uses this information to决定是否 trigger feature recalculation. This feedback loop enables real-time analytics capability while avoiding unnecessary computational resource consumption by only processing when changes are detected.
3Manufacturing precision
If all features are recalculated, then accuracy of updated features is improved, but processing efficiency decreases
Solution Approach 1:
The patent extracts and identifies only the specific volatile features that have changed from the complete feature set. By taking out only the necessary features for recalculation rather than processing all features, the system maintains accuracy for updated features while dramatically improving processing efficiency by excluding unchanged features.
Solution Approach 2:
The patent applies partial action by recalculating only the subset of volatile features that have changed, rather than performing excessive full recalculation of all features. This partial approach maintains necessary accuracy while improving processing efficiency by avoiding redundant computations.
Data Source
AI summary
A system may include a memory and a processor in communication with the memory. The processor may be configured to perform operations. The operations may include receiving data, mapping a plurality of features to a plurality of attributes of the data, and generating a feature-to-attribute map based on said mapping. The operations may further include submitting new data to a feature calculation loop and implementing the feature calculation loop. The operations may also include tracking the features in the feature calculation loop, identifying a changed feature in the features, and recalculating for the changed feature using the feature calculation loop.