Multi-model Time Series Forecasting for Set-Level Variables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods face challenges in accurately and efficiently forecasting future values of a variable that is a function of other values associated with members of a set, particularly in scenarios involving time-varying travel characteristics, where data availability is limited and characteristics do not follow preset rules.
Innovation Solution
A method involving multiple models is employed to forecast future timeseries values, where a first model predicts a set-level variable and one or more second models predict individual variables, with a composite timeseries forecast generated by combining these predictions, optimizing forecast accuracy and reducing computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single model is used to forecast set-level variables, then the device complexity is reduced, but the forecast precision deteriorates
Solution Approach 1:
The patent divides the forecasting task into multiple independent models: a first model forecasts the set-level variable directly, while second models forecast individual member variables. Each model is simpler and more specialized, yet their combined output achieves higher overall forecast precision through the composite timeseries forecast.
Solution Approach 2:
The patent combines the outputs of multiple independent forecasting models into a single composite timeseries forecast. The first future timeseries from the set-level model and second future timeseries from individual member models are merged to produce the final forecast, achieving superior precision without requiring any single model to be overly complex.
2Measurement precision
If multiple models are used to forecast set-level variables, then the forecast precision is improved, but the computational resource usage increases
Solution Approach 1:
The computational task is segmented across multiple specialized models rather than using one monolithic model. Each model handles a specific aspect (set-level or individual member level), reducing the computational burden on each model while maintaining high forecast accuracy through their combined output.
Solution Approach 2:
The individual member variables act as intermediaries between the complex set-level forecasting problem and the final composite forecast. The second models forecast these intermediary variables, which then feed into the composite timeseries forecast, distributing computational work and improving overall efficiency.
3Adaptability or versatility
If traditional forecasting methods are used for set-level variables, then the ease of operation is maintained, but the adaptability to complex scenarios deteriorates
Solution Approach 1:
The forecasting system is segmented into modular components that can be independently configured and operated. The first model and second models can be selected and tuned based on specific scenario requirements, providing adaptability while maintaining operational simplicity through standardized interfaces and composite forecast generation.
Solution Approach 2:
The system dynamically adapts to different forecasting scenarios by selectively applying the first model, second models, or both depending on the specific requirements. The composite timeseries forecast dynamically integrates inputs from different model types, enabling versatility across various complex scenarios while maintaining ease of operation through a unified forecasting framework.
Data Source
AI summary
A method includes predicting a first future timeseries for a first variable using a first model, predicting future values for a plurality of second variables using one or more second models, wherein the first variable is a function of the second variables, generating a second future timeseries for the first variable as based on the future values for the plurality of second variables, and providing a composite timeseries forecast for the first variable by combining the first future timeseries and the second future timeseries.