Time-Series Aggregation With Time-Shifted Event Triggers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Aggregating large volumes of time-series data is computationally intensive and requires efficient algorithms and infrastructure, and handling inconsistent data formats, structures, or schemas, especially when dealing with varying time intervals and outliers, which can skew aggregated values and lose crucial information.
Innovation Solution
The system generates a synthetic historic data set based on a plurality of historic data sets to maintain sequential relationships, determines event-trigger pairings within this synthetic data set, and time-shifts time points to ensure accurate triggers, using artificial intelligence models to predict and respond to events in data-sparse environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simple aggregation methods are used for time-series data, then processing speed is improved, but information loss occurs due to overlooking seasonality and trends
Solution Approach 1:
The patent transforms the aggregation problem by changing the parameter space from direct time-value aggregation to state-space aggregation. By converting time-series data into state representations that encode temporal patterns (seasonality, trends), the system maintains information while enabling efficient aggregation operations. The state space transformation allows aggregation to occur in a different parameter domain where computational efficiency is improved without losing temporal characteristics.
2Adaptability or versatility
If data aggregation is performed with varying time intervals, then adaptability to different data sources is improved, but measurement precision deteriorates due to misalignment and missing values
Solution Approach 1:
The patent introduces state representations as an intermediary layer between raw time-series data with varying intervals and the aggregation operation. This intermediary state space serves as a mediator that normalizes temporal differences, allowing data from different time intervals to be aggregated without direct temporal alignment. The state representation captures essential temporal characteristics while being invariant to sampling frequency, thus preserving measurement precision across diverse time intervals.
3Extent of automation
If aggregation is triggered based on aggregated data thresholds, then automation level is improved, but reliability deteriorates due to delayed triggers from processing lags
Solution Approach 1:
The patent performs preliminary aggregation of state representations before actual triggering decisions are made. By pre-computing aggregated states and storing them, the system eliminates the need for real-time aggregation computation when triggers are evaluated. This preliminary action separates the computationally intensive aggregation step from the trigger decision step, allowing triggers to respond immediately to pre-computed states without suffering from processing lags, thus maintaining both automation and reliability.
Data Source
AI summary
Systems and methods are described herein for novel uses and/or improvements to data aggregation related to artificial intelligence applications, specifically applications related to aggregating time-series data. As one example, systems and methods are described herein for predicting effects when aggregating time-series data and modifying the one or more data streams used to populate a model profile and/or feed an artificial intelligence application with the time-series data.