Auto-Encoder Time Series Forecasting Algorithm Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for selecting forecasting algorithms for time series data are resource-intensive and often inaccurately select algorithms, failing to address the diverse behaviors of time series across various domains.
Innovation Solution
An automated system that trains an auto-encoder to generate time series based on descriptive characteristics, uses this auto-encoder to train a classification model to select the best-suited forecasting model for a given time series, thereby addressing the resource-intensiveness and domain-specific limitations of existing methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing methods are used to select forecasting algorithms, then algorithm selection can be performed, but the process is resource-intensive and may select inaccurate algorithms
Solution Approach 1:
The system pre-trains the auto-encoder model on comprehensive time series data to learn optimal forecasting algorithm selections before actual deployment. This preliminary training phase allows the system to make accurate algorithm selections during runtime without consuming extensive computational resources, as the heavy lifting is done in advance.
Solution Approach 2:
The patent introduces an auto-encoder as an intermediary component that transforms time series data into a compressed representation, which then feeds into the forecasting algorithm selection process. This intermediary layer enables more efficient and accurate algorithm selection by capturing essential patterns without requiring the system to process all raw data features directly.
2Adaptability or versatility
If existing methods are used to select forecasting algorithms, then selection can be made, but they fail to address diverse behaviors of time series across various domains
Solution Approach 1:
The auto-encoder model is designed with universal applicability across multiple time series domains. By training on diverse time series data from various domains during the pre-training phase, the model learns domain-agnostic patterns and characteristics, enabling it to accurately select forecasting algorithms for different types of time series without requiring domain-specific customization or increasing system complexity.
Data Source
AI summary
Systems, methods, and other embodiments associated with characteristics-based selection of time series forecast algorithms are described. In one embodiment, a method includes analyzing each time series in a training set of time series to yield characteristics vectors for the time series. The method trains an auto-encoder to minimize error: (1) between bottleneck layer and the characteristics vectors, and (2) between input layer and output layer. The method generates new characteristics vectors that fill gaps between neighboring characteristics. The method inputs the new characteristics vectors to the bottleneck layer to generate a testing set of time series. The method tests forecasting algorithms using the testing set to find forecasting error. The method trains a ranking function to assign a rank to each forecasting algorithm based on a provided characteristics vector. And, the method automatically selects one of the forecasting algorithms to monitor an additional time series based on the ranks.