Parallel Time Series Forecasting With Best-Fit Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing time series forecasting methods face challenges due to inherent non-stationarity and uncertainty, making it difficult to accurately predict future trends, especially when vast amounts of data are involved, and traditional machine learning models are ineffective for real-time updates.
Innovation Solution
A system and method for time series forecasting that allows simultaneous training and selection of multiple models for each forecast, using auto-regressive integrated moving average models with hyper-parameter tuning, and determining the best fit model based on Akaike information criterion, enabling parallel processing of multiple forecasts with a single query.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional machine learning models are used for time series forecasting, then model training can be performed, but the processing time increases significantly when handling vast amounts of data and multiple forecasts
Solution Approach 1:
The patent segments the forecasting process into independent parallel tasks by dividing multiple time series forecasts into separate processing units. Each forecast is handled independently through vectorized operations, allowing simultaneous computation across multiple data points without sequential bottlenecks.
Solution Approach 2:
The patent replaces traditional iterative machine learning model training with vectorized mathematical operations. Instead of using complex mechanical training processes that require sequential computation, the system uses direct vectorized calculations that can be executed in parallel, dramatically reducing processing time while maintaining forecast accuracy.
2Measurement precision
If multiple models are trained sequentially for each time series forecast, then model accuracy can be improved, but the complexity and time required increases
Solution Approach 1:
The patent merges multiple model training processes into a single unified vectorized operation. Instead of training models separately for each time series, the system combines all training operations into parallel vectorized computations that achieve the same accuracy with reduced complexity and faster execution.
Solution Approach 2:
The patent changes the computational parameters from sequential scalar operations to parallel vector operations. By transforming the mathematical operations into vectorized form, the system maintains model accuracy while reducing the computational complexity from O(n) sequential training to O(1) parallel vectorized training.
3Adaptability or versatility
If traditional forecasting methods are used, then individual forecasts can be generated, but real-time updates and large-scale processing become inefficient
Solution Approach 1:
The patent introduces dynamic parallel processing capabilities that allow the system to adaptively handle varying forecast requirements in real-time. The vectorized operations can dynamically adjust to different data sizes and forecast quantities without retraining, enabling real-time updates while maintaining high processing efficiency for large-scale operations.
Data Source
AI summary
A method for time series forecasting includes receiving a time series forecasting query from a user requesting the data processing hardware to perform a plurality of time series forecasts. Each time series forecast is a forecast of future data based on respective current data. Simultaneously, for each time series forecast of the plurality of time series forecasts requested by the time series forecasting query, the method includes training a plurality of models for the respective time series forecast. The method also includes determining which model of the plurality of models best fits the respective time series forecast and forecasting the future data based on the determined best fitting model and the respective current data. The method also includes returning, to the user, the forecasted future data for each of the plurality of time series forecasts request by the timer series forecasting query.