Time-Series Data Preprocessing for Irregular Intervals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting future health conditions using time-series medical data face challenges due to irregular time intervals and missing values, which reduce the accuracy and reliability of prediction results.
Innovation Solution
A time-series data processing device that includes a preprocessor and a learner. The preprocessor generates time-series interval data and feature interval data to correct irregular time intervals and missing values, while the learner generates a weight group for a prediction model based on the preprocessed data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If time-series medical data with irregular time intervals and missing values is used directly for prediction, then the processing simplicity is maintained, but the prediction accuracy and reliability deteriorate
Solution Approach 1:
The preprocessor performs preliminary actions by generating time-series interval data and feature interval data before the prediction model is applied. This preprocessing step calculates the time intervals between consecutive measurements and creates interval representations of features, preparing the data in advance to handle irregularity and missing values, thereby improving prediction accuracy without requiring complex processing during the prediction phase
Solution Approach 2:
The patent introduces interval data as an intermediary representation between the raw time-series data and the prediction model. By transforming irregular time-series data into interval-based features (time intervals between measurements, feature intervals), the system mediates the gap between irregular input data and the requirements of prediction algorithms, enabling accurate predictions while maintaining processing efficiency
2Reliability
If irregular time intervals and missing values are corrected through preprocessing, then the prediction reliability is improved, but the processing time and computational load increase
Solution Approach 1:
The preprocessor extracts the essential temporal information by taking out the time interval components from the raw time-series data. Instead of complex imputation methods, the system extracts simple interval measurements (time between consecutive observations, feature value intervals) and uses these extracted features for prediction, improving reliability while minimizing processing time
Solution Approach 2:
The patent changes the parameters of the data representation by transforming absolute time stamps and raw feature values into interval-based parameters. By representing data as differences and intervals rather than absolute values, the system handles missing values and irregular timing more efficiently, improving prediction reliability with computationally lightweight transformations
3Measurement precision
If interval data and masking data are generated to handle missing values, then the prediction accuracy is improved, but the data complexity and storage requirements increase
Solution Approach 1:
Instead of generating comprehensive interpolated data for all missing values, the preprocessor applies partial action by creating interval representations that capture the essential information needed for prediction. The masking data selectively identifies missing value positions without requiring full data reconstruction, achieving improved prediction accuracy while minimizing the increase in data volume
Data Source
AI summary
Disclosed is a time-series data processing device that includes a preprocessor, a learner, and a predictor. The preprocessor generates time-series interval data based on a time interval of time-series data, generates feature interval data based on a time interval of each of features of the time-series data, and preprocesses the time-series data. The learner generates a weight group of a prediction model for generating a prediction result based on the time-series interval data, the feature interval data, and the preprocessed time-series data. The predictor generates a time-series weight, which depends on a feature weight of each of the features and a time flow of the time-series data, based on the time-series interval data, the feature interval data, and the preprocessed time-series data and generates a prediction result based on the feature weight and the time-series weight.