Time Series Prediction Using Reference Data Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning systems face challenges in generating predictive models for series with partial or no historical data, as they require sufficient data to forecast future behavior effectively.
Innovation Solution
A system that segments timestamped data to infer performance by detecting the absence or partial presence of data points, using metrics from reference series to impute values and generate predicted performance for series with no or partial history, employing a data processing system with processors and machine learning models to create predictive models from segmented data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learning systems use traditional forecasting methods requiring sufficient historical data, then model accuracy is improved, but the system cannot handle series with partial or no historical data (cold-start problem)
Solution Approach 1:
The patent introduces reference series as intermediary elements to bridge the gap between series with insufficient history and the forecasting model. When a target series has partial or no historical data, the system uses reference series (series with sufficient historical data) as mediators to provide the necessary training information, allowing the model to generate predictions without requiring the target series itself to have extensive history.
Solution Approach 2:
The patent segments the time series data into different categories based on the amount of historical data available (e.g., series with sufficient history, series with partial history, series with no history). This segmentation allows the system to apply different processing strategies to different groups, using reference series for cold-start cases while maintaining traditional methods for well-established series, thus resolving the contradiction between data requirements and model accuracy.
2Adaptability or versatility
If the system segments timestamped data to detect absence or partial presence of data points, then predictive capability for cold-start series is improved, but system complexity increases
Solution Approach 1:
The system performs self-service by automatically detecting the presence or absence of data points and autonomously selecting appropriate reference series without requiring manual intervention. The segmentation process and reference series selection are implemented as automated procedures within the system, reducing the need for external complexity while maintaining high adaptability to different data scenarios.
3Measurement precision
If the system uses reference series to impute values for series with limited history, then prediction accuracy for cold-start series is improved, but computational resources and processing time increase
Solution Approach 1:
The system applies partial action by using only the necessary portion of reference series data required for imputation, rather than processing entire reference series comprehensively. The imputation process selectively uses historical data from reference series only for the time periods and features where the target series has missing or insufficient data, reducing overall computational burden while maintaining prediction accuracy for cold-start series.
Data Source
AI summary
Aspects of this technical solution can segment a first time period for the first series into a second time period bounded by a first time stamp and a second time stamp later than the first time stamp, and into a third time period bounded by a third time stamp later than the second timestamp and a fourth time stamp later than the third time stamp, determine a metric for the third time period and based on first data points of a training data set for the first series and having time stamps bounded by the first time stamp and the second time stamp within the second time period, generate data points within the third time period based on the first metric and generate data points corresponding to a performance of a second series subsequent to the prediction time stamp.