Time series prediction execution based on deviation risk evaluation
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.
1 . A computer-implemented method, comprising:
performing a cross-validation assessment of a plurality of predictive models, wherein the cross-validation assessment is based on time series data that is divided into an estimation data set and a validation data set, and wherein the estimation data set is used as training data;
based on the cross-validation assessment, determining, for each predictive model, a respective deviation risk, wherein the respective deviation risk is determined based on comparing forecasting variability distribution for the validation data set during the cross-validation assessment with forecasting variability distribution for test values from a test data set, wherein the test data set represents forecasted values generated based on a respective predictive model for a future horizon;
excluding one or more predictive models of the plurality of predictive models to define a set of candidate predictive models, wherein the excluding is based on evaluating deviation risks of each of the predictive models of the plurality of predictive models;
performing a model selection of a candidate model from the set of candidate predictive models, wherein the candidate model is selected based on evaluation of accuracy of the set of candidate predictive model according to the cross-validation assessment; and
providing the candidate model for execution of a prediction for a requested time horizon.
2 . The computer-implemented method of claim 1 , comprising:
obtaining the time series data, which comprises data observations associated with a date as a time dimension; and
generating the plurality of predictive models for predicting a measure variable determined from the data observations.
3 . The computer-implemented method of claim 1 , wherein the plurality of predictive models comprises one or more of a double exponential smoothing model, an auto regression model, a linear regression model, and an exponential smoothing model.
4 . The computer-implemented method of claim 1 , comprising:
executing the candidate model to provide an output including predicted values for the requested time horizon.
5 . The computer-implemented method of claim 4 , comprising:
providing the output for use in automation of a service process execution.
6 . The computer-implemented method of claim 1 , wherein the comparing of the forecasting variability distribution is performed based on a deviation rejection rule for excluding predictive models that experience deviation in the forecasting variability distribution above a threshold.
7 . The computer-implemented method of claim 1 , wherein the cross-validation assessment is performed over a plurality of definitions of an estimation data set and a validation set based on respective different cut-off points to divide the time series data at different subsequent time points.
8 . The computer-implemented method of claim 1 , wherein determining a first deviation risk for a first predictive model from the plurality of predictive models comprises:
generating a first test data set for a future test time horizon, the first test data set being generated based on the first predictive model,
wherein the first predictive model is trained based on a first set of estimation data sets of the time series data, and the first predictive model is validated based on a first set of validation data sets of the time series data, wherein each set of the first set of estimation data sets map to a respective set of the first set of validation data sets and to a cut-off point for the time series data.
9 . A non-transitory, computer-readable medium coupled to one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
performing a cross-validation assessment of a plurality of predictive models, wherein the cross-validation assessment is based on time series data that is divided into an estimation data set and a validation data set, and wherein the estimation data set is used as training data;
based on the cross-validation assessment, determining, for each predictive model, a respective deviation risk, wherein the respective deviation risk is determined based on comparing forecasting variability distribution for the validation data set during the cross-validation assessment with forecasting variability distribution for test values from a test data set, wherein the test data set represents forecasted values generated based on a respective predictive model for a future horizon;
excluding one or more predictive models of the plurality of predictive models to define a set of candidate predictive models, wherein the excluding is based on evaluating deviation risks of each of the predictive models of the plurality of predictive models;
performing a model selection of a candidate model from the set of candidate predictive models, wherein the candidate model is selected based on evaluation of accuracy of the set of candidate predictive model according to the cross-validation assessment; and
providing the candidate model for execution of a prediction for a requested time horizon.
10 . The non-transitory computer-readable medium of claim 9 , wherein the non-transitory computer-readable medium stores instructions which when executed by the one or more processors, cause the one or more processors to perform operations comprising:
obtaining the time series data, which comprises data observations associated with a date as a time dimension; and
generating the plurality of predictive models for predicting a measure variable determined from the data observations.
11 . The non-transitory computer-readable medium of claim 9 , wherein the plurality of predictive models comprises one or more of a double exponential smoothing model, an auto regression model, a linear regression model, and an exponential smoothing model.
12 . The computer-implemented method of claim 9 , wherein the non-transitory computer-readable medium stores instructions which when executed by the one or more processors, cause the one or more processors to perform operations comprising:
executing the candidate model to provide an output including predicted values for the requested time horizon.
13 . The non-transitory computer-readable medium of claim 12 , wherein the non-transitory computer-readable medium stores instructions which when executed by the one or more processors, cause the one or more processors to perform operations comprising:
providing the output for use in automation of a service process execution.
14 . The non-transitory computer-readable medium of claim 9 , wherein the comparing of the forecasting variability distribution is performed based on a deviation rejection rule for excluding predictive models that experience deviation in the forecasting variability distribution above a threshold.
15 . A computer-implemented system comprising:
one or more processors; and
one or more computer-readable memories coupled to the one or more processors and having instructions stored thereon that are executable by the one or more processors to perform operations comprising:
performing a cross-validation assessment of a plurality of predictive models, wherein the cross-validation assessment is based on time series data that is divided into an estimation data set and a validation data set, and wherein the estimation data set is used as training data;
based on the cross-validation assessment, determining, for each predictive model, a respective deviation risk, wherein the respective deviation risk is determined based on comparing forecasting variability distribution for the validation data set during the cross-validation assessment with forecasting variability distribution for test values from a test data set, wherein the test data set represents forecasted values generated based on a respective predictive model for a future horizon;
excluding one or more predictive models of the plurality of predictive models to define a set of candidate predictive models, wherein the excluding is based on evaluating deviation risks of each of the predictive models of the plurality of predictive models;
performing a model selection of a candidate model from the set of candidate predictive models, wherein the candidate model is selected based on evaluation of accuracy of the set of candidate predictive model according to the cross-validation assessment; and
providing the candidate model for execution of a prediction for a requested time horizon.
16 . The system of claim 15 , wherein the one or more computer-readable memories stores instructions which when executed by the one or more processors, cause the one or more processors to perform operations comprising:
obtaining the time series data, which comprises data observations associated with a date as a time dimension; and
generating the plurality of predictive models for predicting a measure variable determined from the data observations.
17 . The system of claim 15 , wherein the plurality of predictive models comprises one or more of a double exponential smoothing model, an auto regression model, a linear regression model, and an exponential smoothing model.
18 . The system of claim 15 , wherein the one or more computer-readable memories stores instructions which when executed by the one or more processors, cause the one or more processors to perform operations comprising:
executing the candidate model to provide an output including predicted values for the requested time horizon.
19 . The system of claim 18 , wherein the one or more computer-readable memories stores instructions which when executed by the one or more processors, cause the one or more processors to perform operations comprising:
providing the output for use in automation of a service process execution.
20 . The system of claim 15 , wherein the comparing of the forecasting variability distribution is performed based on a deviation rejection rule for excluding predictive models that experience deviation in the forecasting variability distribution above a threshold.