Automatic machine learning feature backward stripping
Features are used to train one or more ML models in a modelling layer. In a feature selection layer, each generated ML model is analyzed to determine, for each input feature, a degree of importance of the feature on the results generated by the ML model. Features with low importance are identified and the information is propagated backward to the data source and feature engineering layers. In response, the data source and feature engineering layers refrain from gathering or generating the unimportant features. Based on a confidence measure of the determination that each feature is important or unimportant, a number of periods between reevaluation of the feature importance is determined. After the number of periods has elapsed, a removed feature is restored to the pipeline.
1 . A method comprising:
training, by one or more processors, a first set of machine learning (ML) models based on a current subset of a set of features;
determining, by the one or more processors, that a feature of the current subset of features has low importance;
based on the low importance of the feature, modifying the current subset of the set of features by removing the feature;
determining, for the feature, a confidence measure in the low importance of the feature;
based on the confidence measure, determining a number of iterations to delay reintroduction of the feature to the current subset of the set of features; and
training a second set of ML models based on the modified current subset of the set of features.
2 . The method of claim 1 , wherein the determining of the confidence measure is based on a fraction of the first set of ML models in which the feature was important.
3 . The method of claim 1 , wherein the number of iterations to delay reintroduction of the feature is larger when the confidence measure is larger.
4 . The method of claim 1 , wherein the training of the first set of ML models comprises training a set of random forest models.
5 . The method of claim 1 , wherein the training of the first set of ML models comprises training a set of gradient boosting decision trees.
6 . The method of claim 1 , wherein the first set of ML models are generated at a predetermined rate.
7 . The method of claim 1 , wherein the set of features comprises raw data and engineered data.
8 . A system comprising:
a memory that stores instructions; and
one or more processors configured by the instructions to perform operations comprising:
training a first set of machine learning (ML) models based on a current subset of a set of features;
determining that a feature of the current subset of features has low importance;
based on the low importance of the feature, modifying the current subset of the set of features by removing the feature;
determining, for the feature, a confidence measure in the low importance of the feature;
based on the confidence measure, determining a number of iterations to delay reintroduction of the feature to the current subset of the set of features; and
training a second set of ML models based on the modified current subset of the set of features.
9 . The system of claim 8 , wherein the determining of the confidence measure is based on a fraction of the first set of ML models in which the feature was important.
10 . The system of claim 8 , wherein the number of iterations to delay reintroduction of the feature is larger when the confidence measure is larger.
11 . The system of claim 8 , wherein the training of the first set of ML models comprises training a set of random forest models.
12 . The system of claim 8 , wherein the training of the first set of ML models comprises training a set of gradient boosting decision trees.
13 . The system of claim 8 , wherein the first set of ML models are generated at a predetermined rate.
14 . The system of claim 8 , wherein the set of features comprises raw data and engineered data.
15 . A non-transitory machine-readable medium that stores instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
training a first set of machine learning (ML) models based on a current subset of a set of features;
determining that a feature of the current subset of features has low importance;
based on the low importance of the feature, modifying the current subset of the set of features by removing the feature;
determining, for the feature, a confidence measure in the low importance of the feature;
based on the confidence measure, determining a number of iterations to delay reintroduction of the feature to the current subset of the set of features; and
training a second set of ML models based on the modified current subset of the set of features.
16 . The non-transitory machine-readable medium of claim 15 , wherein the determining of the confidence measure is based on a fraction of the first set of ML models in which the feature was important.
17 . The non-transitory machine-readable medium of claim 15 , wherein the number of iterations to delay reintroduction of the feature is larger when the confidence measure is larger.
18 . The non-transitory machine-readable medium of claim 15 , wherein the training of the first set of ML models comprises training a set of random forest models.
19 . The non-transitory machine-readable medium of claim 15 , wherein the training of the first set of ML models comprises training a set of gradient boosting decision trees.
20 . The non-transitory machine-readable medium of claim 15 , wherein the first set of ML models are generated at a predetermined rate.