Learning Device State Vector Generation for Time-Series Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting future values from time-series data, such as those using recurrent neural networks, lack accuracy in capturing complex trends and fluctuations, particularly in demand prediction for service parts.
Innovation Solution
A learning device and method that generate a state vector from time-series data by calculating differences and powers, and perform learning using dynamic mode decomposition and regression models to improve prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional recurrent neural networks are used for time-series prediction, then the system can process sequential data, but the prediction accuracy for complex trends and fluctuations is insufficient
Solution Approach 1:
The patent segments the time-series data into trend components and residual components through decomposition. This segmentation allows the model to separately analyze and predict different characteristics of the data, improving overall prediction accuracy by handling complex patterns in a structured manner
Solution Approach 2:
The patent transforms the input data by calculating differences between consecutive time points and computing power terms (squared values). These parameter transformations enhance the model's ability to capture temporal dynamics and non-linear patterns, thereby improving prediction accuracy for complex trends
2Measurement precision
If more complex models are used to capture complex trends, then prediction accuracy improves, but computational cost and data requirements increase
Solution Approach 1:
By transforming raw data into difference sequences and power terms, the model extracts meaningful features that capture temporal patterns more efficiently. This parameter transformation allows the model to achieve high prediction accuracy with fewer data points, reducing the quantity of substance requirement
Data Source
AI summary
A learning device is configured to generate from time-series data a state vector including a first component including a difference between data at a target time and data at a time earlier than the target time and a second component including a power of the data at the target time, and perform learning using the state vector.