Automated machine learning system
A technology is described for automating deployment of a machine learning model. An example method may include receiving, via a graphical user interface, credentials for connecting to a data store containing a plurality of datasets and connecting to the data store using the credentials. A selection of a target metric to predict using the machine learning model can be received, via the graphical user interface, and datasets included in the plurality of datasets that correlate to the target metric can be identified by analyzing the datasets to identify an association between the target metric and data contained within the datasets. The datasets can be input to the machine learning model to train the machine learning model to generate predictions of the target metric, and the machine learning model can be deployed to computing resources in a service provider environment to generate predictions associated with the target metric.
1. An apparatus comprising:
at least one processor; and
memory storing processor-executable instructions that, when executed by the at least one processor, cause the apparatus to:
receive, via a graphical user interface (GUI) output at a client device, a selection of a target metric to predict, wherein the GUI comprises at least one input control that facilitates adjusting an amount of influence for a prediction driver associated with the target metric, wherein the client device is in communication with a user data store and a service provider environment;
determine, based on the selection of the target metric, one or more datasets of a plurality of datasets comprising data records storing values based on the prediction driver, wherein the plurality of data sets are stored at the user data store;
cause, based on the one or more datasets and the prediction driver, a first version of a machine learning model to be generated and a first plurality of predictions for the target metric to be generated by the first version of the machine learning model;
receive, via the at least one input control, an adjustment to the amount of influence for the prediction driver, wherein the adjustment causes at least one value stored in the data records within the one or more datasets to be modified with a substitute value, wherein the substitute value is based on the at least one value and the adjustment;
cause, based on the modified at least one value within the one or more datasets, at least one further version of the machine learning model to be generated and at least one further plurality of predictions for the target metric to be generated by the at least one further version of the machine learning model;
output, via the GUI, a first prediction score for the first version of the machine learning model and at least one further prediction score for the at least one further version of the machine learning model, wherein the first prediction score is based on the first plurality of predictions for the target metric, and wherein the at least one further prediction score is based on the at least one further plurality of predictions for the target metric;
receive, via the GUI, a selection of a version of the machine learning model for deployment, wherein the selection is made following the output of the first prediction score and the at least one further prediction score; and
based on the selection, cause:
the version of the machine learning model selected for deployment to be deployed to computing resources in the service provider environment, and
data associated with the at least one further plurality of predictions and the modified at least one value within the one or more datasets to be stored in the user data store.
2. The apparatus of claim 1 , wherein the at least one input control comprises a slider control, an increment and decrement control, or an input box.
3. The apparatus of claim 1 , wherein the adjustment to the amount of influence for the prediction driver comprises an adjustment to a weight or a degree of influence associated with the prediction driver.
4. The apparatus of claim 1 , wherein the processor-executable instructions further cause the apparatus to:
receive, via the GUI, a selection of a data element in the one or more datasets identified as not useful to predict the target metric; and
remove the data element from the one or more datasets.
5. The apparatus of claim 1 , wherein the processor-executable instructions further cause the apparatus to:
modify the at least one value within the one or more datasets associated with the prediction driver in proportion to the adjustment to the amount of influence for the prediction driver.
6. The apparatus of claim 1 , wherein the processor-executable instructions further cause the apparatus to:
determine, based on the first plurality of predictions for the target metric, the first prediction score, wherein the first prediction score is indicative of performance of the first version of the machine learning model to predict the target metric; and
determine, based on the at least one further plurality of predictions for the target metric, the at least one further prediction score, wherein the at least one further prediction score is indicative of performance of the at least one further version of the machine learning model to predict the target metric.
7. The apparatus of claim 1 , wherein the processor-executable instructions further cause the apparatus to:
train the first version of the machine learning model using the one or more datasets; and
train the at least one further version of the machine learning model using the modified at least one value within the one or more datasets.
8. A method, comprising:
receiving, via a graphical user interface (GUI) output at a client device, a selection of a target metric to predict, wherein the GUI comprises at least one input control that facilitates adjusting an amount of influence for a prediction driver associated with the target metric, wherein the client device is in communication with a user data store and a service provider environment;
determining, based on the selection of the target metric, one or more datasets of a plurality of datasets; comprising data records storing values based on the prediction driver, wherein the plurality of data sets are stored at the user data store;
causing, based on the one or more datasets and the prediction driver, a first version of a machine learning model to be generated and a first plurality of predictions for the target metric to be generated by the first version of the machine learning model;
receiving, via the at least one input control, an adjustment to the amount of influence for the prediction driver, wherein the adjustment causes at least one value stored in the data records within the one or more datasets to be modified with a substitute value, wherein the substitute value is based on the at least one value and the adjustment;
causing, based on the modified at least one value within the one or more datasets, at least one further version of the machine learning model to be generated and at least one further second plurality of predictions for the target metric to be generated by the at least one further version of the machine learning model;
outputting, via the GUI, a first prediction score for the first version of the machine learning model and at least one further prediction score for the at least one further version of the machine learning model, wherein the first prediction score is based on the first plurality of predictions for the target metric, and wherein the at least one further prediction score is based on the at least one further plurality of predictions for the target metric;
receiving, via the GUI, a selection of a version of the machine learning model for deployment, wherein the selection is made following the output of the first prediction score and the at least one further prediction score; and
based on the selection, causing:
the version of the machine learning model selected for deployment to be deployed to computing resources in the service provider environment, and
data associated with the at least one further plurality of predictions and the modified at least one value within the one or more datasets to be stored in the user data store.
9. The method of claim 8 , wherein the at least one input control comprises a slider control, an increment and decrement control, or an input box.
10. The method of claim 8 , wherein the adjustment to the amount of influence for the prediction driver comprises an adjustment to a weight or a degree of influence associated with the prediction driver.
11. The method of claim 8 , further comprising: modifying the at least one value within the one or more datasets associated with the prediction driver in proportion to the adjustment to the amount of influence for the prediction driver.
12. The method of claim 8 , further comprising:
determining, based on the first plurality of predictions for the target metric, the first prediction score, wherein the first prediction score is indicative of performance of the first version of the machine learning model to predict the target metric; and
determining, based on the at least one further plurality of predictions for the target metric, the at least one further prediction score, wherein the at least one further prediction score is indicative of performance of the at least one further version of the machine learning model to predict the target metric.
13. The method of claim 8 , further comprising:
training the first version of the machine learning model using the one or more datasets; and
training the at least one further version of the machine learning model using the modified at least one value within the one or more datasets.
14. A non-transitory computer-readable storage medium storing processor-executable instructions that, when executed by one or more processors of a computing device, cause the computing device to:
receive, via a graphical user interface (GUI) output at a client device, a selection of a target metric to predict, wherein the GUI comprises at least one input control that facilitates adjusting an amount of influence for a prediction driver associated with the target metric, wherein the client device is in communication with a user data store and a service provider environment;
determine, based on the selection of the target metric, one or more datasets of a plurality of datasets comprising data records storing values based on the prediction driver, wherein the plurality of data sets are stored at the user data store;
cause, based on the one or more datasets and the prediction driver, a first version of a machine learning model to be generated and a first plurality of predictions for the target metric to be generated by the first version of the machine learning model;
receive, via the at least one input control, an adjustment to the amount of influence for the prediction driver, wherein the adjustment causes at least one value stored in the data records within the one or more datasets to be modified with a substitute value, wherein the substitute value is based on the at least one value and the adjustment;
cause, based on the modified at least one value within the one or more datasets, at least one further version of the machine learning model to be generated and at least one further plurality of predictions for the target metric to be generated by the at least one further version of the machine learning model;
output, via the GUI, a first prediction score for the first version of the machine learning model and at least one further prediction score for the at least one further version of the machine learning model, wherein the first prediction score is based on the first plurality of predictions for the target metric, and wherein the at least one further prediction score is based on the at least one further plurality of predictions for the target metric;
receive, via the GUI, a selection of a version of the machine learning model for deployment, wherein the selection is made following the output of the first prediction score and the at least one further prediction score; and
based on the selection, cause:
the version of the machine learning model selected for deployment to be deployed to computing resources in the service provider environment, and
data associated with the at least one further plurality of predictions and the modified at least one value within the one or more datasets to be stored in the user data store.
15. The non-transitory computer-readable storage medium of claim 14 , wherein the at least one input control comprises a slider control, an increment and decrement control, or an input box.
16. The non-transitory computer-readable storage medium of claim 14 , wherein the adjustment to the amount of influence for the prediction driver comprises an adjustment to a weight or a degree of influence associated with the prediction driver.
17. The non-transitory computer-readable storage medium of claim 14 , wherein the processor-executable instructions further cause the computing device to:
receive, via the GUI, a selection of a data element in the one or more datasets identified as not useful to predict the target metric; and
remove the data element from the one or more datasets.
18. The non-transitory computer-readable storage medium of claim 14 , wherein the processor-executable instructions further cause the computing device to: modify the at least one value within the one or more datasets associated with the prediction driver in proportion to the adjustment to the amount of influence for the prediction driver.
19. The non-transitory computer-readable storage medium of claim 14 , wherein the processor-executable instructions further cause the computing device to:
determine, based on the first plurality of predictions for the target metric, the first prediction score, wherein the first prediction score is indicative of performance of the first version of the machine learning model to predict the target metric; and
determine, based on the at least one further plurality of predictions for the target metric, the at least one further prediction score, wherein the at least one further prediction score is indicative of performance of the at least one further version of the machine learning model to predict the target metric.
20. The non-transitory computer-readable storage medium of claim 14 , wherein the processor-executable instructions further cause the computing device to:
train the first version of the machine learning model using the one or more datasets; and
train the at least one further version of the machine learning model using the modified at least one value within the one or more datasets.