Time Series Model Selection via Deviation Risk Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing time series forecasting models often fail to accurately predict future values due to abrupt changes in data trends, particularly at the end of the time series, leading to significant deviations in forecasting variability, which can result in inaccurate predictions and inefficiencies.
Innovation Solution
A method for selecting a predictive model from a set of candidates based on cross-validation assessment, where models with high deviation risk are excluded by comparing forecasting variability distributions between validation and test data sets, using a deviation rejection rule to filter out models that deviate significantly from the training data, thereby reducing the risk of selecting models that fail to capture late changes in the time series.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional cross-validation is used to select predictive models, then model selection process is simple, but forecasting accuracy deteriorates due to abrupt changes in data trends
Solution Approach 1:
The patent applies preliminary action by performing a preliminary assessment of deviation risk for each candidate model before final model selection. This involves comparing forecasting variability distributions from validation data sets against test data sets to identify and exclude models with high deviation risk, ensuring that only models robust to abrupt data changes proceed to final selection based on accuracy metrics.
Solution Approach 2:
The patent segments the model selection process into distinct phases: (1) initial candidate model generation, (2) deviation risk assessment and exclusion, (3) cross-validation of remaining candidates, and (4) final model selection. This segmentation allows systematic handling of data trend changes by filtering models based on their stability characteristics before evaluating their predictive accuracy.
2Reliability
If models are selected based solely on cross-validation accuracy, then selection process is straightforward, but reliability deteriorates due to high deviation risk from late changes in time series
Solution Approach 1:
The patent performs preliminary deviation risk assessment for each candidate model by comparing forecasting variability distributions between validation and test data sets. Models exhibiting high deviation risk are excluded before final selection, ensuring that only reliable models with stable forecasting behavior under varying data conditions proceed to the selection stage.
Solution Approach 2:
The patent establishes equipotentiality by creating a standardized deviation risk assessment framework that uniformly evaluates all candidate models against the same criteria (forecasting variability distribution comparison). This ensures fair and consistent reliability assessment across different model types and structures.
3Measurement precision
If traditional model selection is used, then computational resources are saved, but forecasting precision deteriorates due to inclusion of high-deviation models
Solution Approach 1:
The patent performs preliminary deviation risk assessment by comparing forecasting variability distributions from validation data sets against test data sets. This preliminary filtering excludes models likely to produce inaccurate predictions due to high sensitivity to data changes, thereby improving prediction precision before final model selection and reducing wasted computational resources on poor-performing models.
Data Source
AI summary
The present disclosure relates to computer-implemented methods, software, and systems for identifying data patterns based on data observations collected as time series data. A cross-validation assessment of a plurality of predictive models is performed. Based on the cross-validation assessment, a respective deviation risk is determined. The respective deviation risk is determined based on comparing forecasting variability distribution for a validation data set during the cross-validation assessment with forecasting variability distribution for test values from a test data set. The test data set represents forecasted values generated based on a respective predictive model for a future horizon. A predictive model can be excluded based on evaluating deviation risks of each of the predictive models. A model selection of a candidate model from the set of candidate predictive models is performed. The candidate model is selected based on evaluation of accuracy of the set of candidate predictive model according to the cross-validation assessment.