Artificial intelligence optimization platform
Systems and methods are provided for reusing machine learning models. For example, the applicability of prior models may be compared using one or more assessment values, including a similarity threshold and/or an accuracy threshold. The similarity threshold may identify a similarity of data between a first data set used to generate a first model and a new data set that is received by the system. When the similarity between these two data sets is exceeded, the system may reuse a model with the highest similarity value. When an accuracy value of the data set does not exceed an accuracy threshold, the system may initiate a retraining process to generate a second ML model associated with the second data.
1. A computing platform comprising:
a set of edge devices comprising a first edge device and a second edge device that generate time series data sets in near real time,
wherein the first edge device is configured to:
process a first processing workload in a first environment where the first edge device is located, and
generate a first time series data set associated with the first processing workload;
wherein the second edge device is configured to:
process a second processing workload in a second environment where the second edge device is located, and
generate a second time series data set associated with the second processing workload; and
a central computing system comprising:
a memory; and
one or more processors configured to access the memory and execute machine readable instructions stored to:
receive a first machine learning (ML) model, the first ML model being trained using the first time series data set;
receive the second time series data set, the second time series data set being generated by the second edge device;
determine that the second time series data set comprises a data characteristic;
compare the second time series data set with the first time series data set to generate a similarity value associated with the data characteristic;
when the similarity value exceeds a similarity threshold:
reuse the first ML model to process the second time series data set,
generate an output from the first ML model, and
measure an accuracy value based on the output;
compare the accuracy value to an accuracy threshold; and
when the accuracy value does not exceed the accuracy threshold or when the similarity value does not exceed the similarity threshold, initiate a retraining process to generate a second ML model associated with the second time series data set.
2. The computing platform of claim 1 , wherein the instructions further to:
when the accuracy value exceeds the accuracy threshold, run forecasting predictions with the first or second ML model.
3. The computing platform of claim 1 , wherein when the accuracy value does not exceed the accuracy threshold, the similarity threshold is adjusted for the first ML model.
4. The computing platform of claim 1 , wherein the similarity threshold and the accuracy threshold are different values.
5. The computing platform of claim 1 , wherein the accuracy value exceeds the accuracy threshold based on a drift-based determination.
6. The computing platform of claim 1 , wherein the instructions further to:
store the second ML model with the first ML model in a model data store.
7. The computing platform of claim 1 , wherein a determination to reuse the first ML model to process the second time series data set is executed in near real time.
8. The computing platform of claim 1 , wherein the first time series data set and the second time series data set are transmitted from the first edge device and the second edge device, respectively, as data matrices where rows represent time instances and columns represent a dimension being ingested.
9. A computer-implemented method comprising:
processing, by a first edge device of a computing platform, a first processing workload in a first environment where the first edge device is located;
generating, by the first edge device, a first time series data set associated with the first processing workload;
processing, by a second edge device of the computing platform, a second processing workload in a second environment where the second edge device is located;
generating, by the second edge device, a second time series data set associated with the second processing workload;
receiving, by a central computing system of a computing platform, a first machine learning (ML) model, the first ML model being trained using the first time series data set;
receiving, by the central computing system, the second time series data set, the second time series data set being generated by the second edge device;
determining, by the central computing system, that the second time series data set comprises a data characteristic;
comparing the second time series data set with the first time series data set to generate an assessment value associated with the data characteristic;
when the assessment value exceeds a similarity threshold:
reusing the first ML model to process the second time series data set,
generating an output from the first ML model, and
measuring an accuracy value based on the output;
comparing the accuracy value to an accuracy threshold; and
when the accuracy value does not exceed the accuracy threshold or when the assessment value does not exceed the similarity threshold, initiating a retraining process to generate a second ML model associated with the second time series data set.
10. The computer-implemented method of claim 9 , further comprising:
when the accuracy value exceeds the accuracy threshold, run forecasting predictions with the first or second ML model.
11. The computer-implemented method of claim 9 , wherein the first data set and the second data set are time series data sets.
12. The computer-implemented method of claim 9 , wherein the similarity threshold and the accuracy threshold are different values.
13. The computer-implemented method of claim 9 , wherein the first data set and the second data set originate from different sources.
14. The computer-implemented method of claim 9 , wherein the accuracy value exceeds the accuracy threshold based on a drift-based determination.
15. The computer-implemented method of claim 9 , further comprising:
storing the second ML model with the first ML model in a model data store.
16. A non-transitory computer-readable storage medium storing a plurality of instructions executable by one or more processors, the plurality of instructions when executed by the one or more processors cause the one or more processors to:
receive a second data set comprising a data characteristic, the second data set being generated by a second edge device or node that collects the second data set from a second processing workload from a second environment where the second edge device is located;
compare the second data set with a first data set to generate a similarity value associated with the data characteristic, the first data set being generated by a first edge device or node that collects the first data set from a first processing workload from a first environment where the first edge device is located, wherein the first data set was used to generate a first machine learning (ML) model;
when the similarity value exceeds a similarity threshold:
reuse the first ML model to process the second data set,
generate an output from the first ML model, and
measure an accuracy value based on the output;
compare the accuracy value to an accuracy threshold; and
when the accuracy value does not exceed the accuracy threshold or when the similarity value does not exceed the similarity threshold, initiate a retraining process to generate a second ML model associated with the second data.
17. The computer-readable storage medium of claim 16 , wherein the plurality of instructions further cause the one or more processors to:
when the accuracy value exceeds the accuracy threshold, run forecasting predictions with the first or second ML model.
18. The computer-readable storage medium of claim 16 , wherein the similarity threshold and the accuracy threshold are different values.
19. The computer-readable storage medium of claim 16 , wherein the first data set and the second data set originate from different sources.
20. The computer-readable storage medium of claim 16 , wherein the accuracy value exceeds the accuracy threshold based on a drift-based determination.