Auto-selection of hierarchically-related near-term forecasting models
Techniques are disclosed herein for selecting a predictive model to perform on a set of hierarchical data. A selection of first time series data representing activity observed in a current period in a first hierarchy is received. Second time series data representing activity observed in a prior period in the first hierarchy is retrieved. Predictive models are performed using the second time series data as input, where each predictive model generates statistical outcomes for the current period. A score is generated in each of the plurality of predictive models based on a comparison of the statistical outcomes with the first time series data. Each of the predictive models is ranked based on the generated scores.
1. A method for selecting a predictive model to perform on a set of hierarchical data, the method comprising:
receiving, via a graphical user interface, a request from a user for a set of forecasted outcomes specifying:
a first time series data, wherein the first time series data represents activity observed in a current period in a first hierarchy and includes metadata with domain-specific knowledge of the first time series data, and
an amount of time for projecting the set of forecasted outcomes;
retrieving a second time series data, wherein the second time series data represents activity observed in at least a prior period of time in the first hierarchy and includes metadata with domain-specific knowledge of the second time series data;
for each predictive model of a plurality of predictive models:
generating statistical outcomes using the second time series data as input;
generating a score for each predictive model of the plurality of predictive models based on a comparison function to compare:
the statistical outcomes generated using the second time series data as input, and
actual outcomes in the first hierarchy corresponding to a same amount of time as the amount of time for projecting the set of forecasted outcomes;
adjusting the generated score for each of the plurality of predictive models based on:
determining based on the metadata with domain-specific knowledge of the second time series data:
a third time series data in a second hierarchy that was previously evaluated by the predictive model; and
effectiveness of the predictive model in generating a score for the third time series data in the second hierarchy;
ranking each predictive model of the plurality of predictive models based on the adjusted score for each of the plurality of predictive models;
generating a set of graphs indicating how each predictive model of the plurality of predictive models performed relative to one another, wherein each graph corresponds to a predictive model of the plurality of predictive models;
selecting a predictive model from the ranking of the plurality of predictive models that has a highest score;
presenting the selected predictive model and a corresponding graph from the set of graphs via the graphical user interface;
receiving confirmation from the user via the graphical user interface to perform the selected predictive model;
applying the selected predictive model to the first time series data; and
generating the set of forecasted outcomes based on the selected predictive model with the first time series data.
2. The method of claim 1 , wherein the first time series data includes contextual information describing the first hierarchy.
3. The method of claim 2 , further comprising: augmenting the score in each predictive model of the plurality of predictive models based on the contextual information.
4. The method of claim 2 , wherein the contextual information describes a performance of one of the plurality of predictive models relative to the second hierarchy.
5. The method of claim 4 , further comprising: augmenting the score of the one of the plurality of predictive models based on the contextual information.
6. A computer-readable storage medium storing instructions, which, when executed on one or more processors, performs an operation for selecting a predictive model to perform on a set of hierarchical data, the operation comprising:
receiving, via a graphical user interface, a request from a user for a set of forecasted outcomes specifying:
a first time series data, wherein the first time series data represents activity observed in a current period in a first hierarchy and includes metadata with domain-specific knowledge of the first time series data, and
an amount of time for projecting the set of forecasted outcomes;
retrieving a second time series data, wherein the second time series data represents activity observed in at least a prior period of time in the first hierarchy and includes metadata with domain-specific knowledge of the second time series data;
for each predictive model of a plurality of predictive models:
generating statistical outcomes using the second time series data as input;
generating a score for each predictive model of the plurality of predictive models based on a comparison function to compare:
the statistical outcomes generated using the second time series data as input, and
actual outcomes in the first hierarchy corresponding to a same amount of time as the amount of time for projecting the set of forecasted outcomes;
adjusting the generated score for each of the plurality of predictive models based on:
determining based on the metadata with domain-specific knowledge of the second time series data:
a third time series data in a second hierarchy that was previously evaluated by the predictive model; and
effectiveness of the predictive model in generating a score for the third time series data in the second hierarchy;
ranking each predictive model of the plurality of predictive models based on the adjusted score for each of the plurality of predictive models;
generating a set of graphs indicating how each predictive model of the plurality of predictive models performed relative to one another, wherein each graph corresponds to a predictive model of the plurality of predictive models;
selecting a predictive model from the ranking of the plurality of predictive models that has a highest score;
presenting the selected predictive model and a corresponding graph from the set of graphs via the graphical user interface;
receiving confirmation from the user via the graphical user interface to perform the selected predictive model;
applying the selected predictive model to the first time series data; and
generating the set of forecasted outcomes based on the selected predictive model with the first time series data.
7. The computer-readable storage medium of claim 6 , wherein the first time series data includes contextual information describing the first hierarchy.
8. The computer-readable storage medium of claim 7 , wherein the operation further comprises: augmenting the score in each predictive model of the plurality of predictive models based on the contextual information.
9. The computer-readable storage medium of claim 7 , wherein the contextual information describes a performance of one of the plurality of predictive models relative to the second hierarchy.
10. The computer-readable storage medium of claim 9 , wherein the operation further comprises: augmenting the score of the one of the plurality of predictive models based on the contextual information.
11. A system, comprising:
one or more processors; and
a memory storing program code, which, when executed on the one or more processors, performs an operation for selecting a predictive model to perform on a set of hierarchical data, the operation comprising:
receiving, via a graphical user interface, a request from a user for a set of forecasted outcomes specifying:
a first time series data, wherein the first time series data represents activity observed in a current period in a first hierarchy and includes metadata with domain-specific knowledge of the first time series data, and
an amount of time for projecting the set of forecasted outcomes;
retrieving a second time series data, wherein the second time series data represents activity observed in at least a prior period of time in the first hierarchy and includes metadata with domain-specific knowledge of the second time series data,
for each predictive model of a plurality of predictive models:
generating statistical outcomes using the second time series data as input,
generating a score for each predictive model of the plurality of predictive models based on a comparison function to compare:
the statistical outcomes generated using the second time series data as input, and
actual outcomes in the first hierarchy corresponding to a same amount of time as the amount of time for projecting the set of forecasted outcomes;
adjusting the generated score for each of the plurality of predictive models based on:
determining based on the metadata with domain-specific knowledge of the second time series data:
a third time series data in a second hierarchy that was previously evaluated by the predictive model; and
effectiveness of the predictive model in generating a score for the third time series data in the second hierarchy;
ranking each predictive model of the plurality of predictive models based on the adjusted score for each of the plurality of predictive models;
generating a set of graphs indicating how each predictive model of the plurality of predictive models performed relative to one another, wherein each graph corresponds to a predictive model of the plurality of predictive models;
selecting a predictive model from the ranking of the plurality of predictive models that has a highest score;
presenting the selected predictive model and a corresponding graph from the set of graphs via the graphical user interface;
receiving confirmation from the user via the graphical user interface to perform the selected predictive model;
applying the selected predictive model to the first time series data; and
generating the set of forecasted outcomes based on the selected predictive model with the first time series data.
12. The system of claim 11 , wherein the first time series data includes contextual information describing the first hierarchy.
13. The system of claim 12 , wherein the operation further comprises:
augmenting the score in each predictive model of the plurality of predictive models based on the contextual information.
14. The system of claim 12 , wherein the contextual information describes a performance of one of the plurality of predictive models relative to the second hierarchy.
15. The system of claim 14 , wherein the operation further comprises: augmenting the score of the one of the plurality of predictive models based on the contextual information.
16. The method of claim 1 , wherein the method further comprises:
determining a periodic time interval; and
generating a package of time series data based on the periodic time interval for the second time series data.
17. The method of claim 16 , wherein the method further comprises:
inputting each package of time series data to each predictive model of the plurality of predictive models.
18. The computer-readable storage medium of claim 6 , wherein the operation comprises:
determining a periodic time interval; and
generating a package of time series data based on the periodic time interval for the second time series data.
19. The computer-readable storage medium of claim 18 , wherein the operation comprises: inputting each package of time series data to each predictive model of the plurality of predictive models.
20. The system of claim 11 , wherein the operation comprises:
determining a periodic time interval; and
generating a package of time series data based on the periodic time interval for the second time series data to input into each predictive model of the plurality of predictive models.