Chained feature synthesis and dimensional reduction

1 . A system comprising:

one or more processors and one or more non-transitory media storing instructions that, when executed, cause the one or more processors to perform operations comprising:

obtaining a first dataset that comprise a first feature set;

generating a first set of feature values by providing the first dataset to a first set of feature primitive stacks;

determining a reduced set of feature values based on the first set of feature values by dimensionally reducing features of the first set of feature values;

generating an intermediate set of feature values by providing a value of the first dataset and a value of the reduced set of feature values to at least one feature primitive of the first set of feature primitive stacks;

updating the reduced set of feature values by dimensionally reducing features of the intermediate set of feature values;

storing a second dataset comprising features of the intermediate set of feature values in association with the first feature set; and

storing a sequence of identifiers in association with the first feature set, the sequence of identifiers comprising (i) a first set of identifiers identifying the first set of feature primitive stacks; (ii) a second set of identifiers identifying a first set of dimension-reducing operators used to determine the reduced set of feature values; (iii) a third set of identifiers identifying a second set of feature primitive stacks used to generate the intermediate set of feature values; and (iv) a fourth set of identifiers identifying a second set of dimension-reducing operators used to dimensionally reduce features of the intermediate set of feature values.

2 . The system of claim 1 , the operations further comprising:

based on a determination that a third dataset comprises the first feature set, retrieving the sequence of identifiers; and

generating a fourth dataset comprising the first feature set based on the sequence of identifiers.

3 . The system of claim 1 , the operations further comprising:

obtaining a set of feature primitives; and

generating the first set of feature primitive stacks by generating sequences of operators based on the set of feature primitives.

4 . The system of claim 1 , wherein providing the first dataset to the first set of feature primitive stacks comprises:

determining whether a first feature of the first feature set was previously provided to a first feature primitive of the first set of feature primitive stacks; and

in response to a determination that the first feature of the first feature set was previously provided to the first feature primitive, retrieving a stored set of feature values associated with the first feature primitive, wherein the first set of feature values comprises the stored set of feature values.

5 . A method comprising:

obtaining a first dataset comprising a first feature set;

generating a first set of feature values by providing the first dataset to a set of feature primitive stacks, wherein providing the first dataset to the set of feature primitive stacks comprises:

determining whether a first feature of the first feature set was previously provided to a first feature primitive of the set of feature primitive stacks; and

in response to a determination that the first feature of the first feature set was previously provided to the first feature primitive, retrieving a stored set of feature values associated with the first feature primitive, wherein the first set of feature values comprises the stored set of feature values;

determining a reduced set of feature values based on the first set of feature values by dimensionally reducing features of the first set of feature values;

generating an intermediate set of feature values by providing a value of the first dataset and a value of the reduced set of feature values to at least one feature primitive of the set of feature primitive stacks;

updating the reduced set of feature values by dimensionally reducing features of the intermediate set of feature values; and

storing a second dataset comprising features of the intermediate set of feature values in association with the first feature set.

6 . The method of claim 5 , further comprising updating the first dataset to remove an outlier of the first dataset by:

generating a set of random decision trees;

determining a set of isolation numbers for an outlier data object of the first dataset based on the set of random decision trees; and

filtering the first dataset to remove the outlier data object from the first dataset.

7 . The method of claim 5 , wherein determining the reduced set of feature values comprises:

obtaining a user-provided feature reduction parameter; and

dimensionally reducing the first set of feature values based on the user-provided feature reduction parameter.

8 . The method of claim 5 , wherein updating the reduced set of feature values comprises performing a set of linear operations based on an additional dataset derived from providing the reduced set of feature values to the set of feature primitive stacks.

9 . The method of claim 5 , wherein updating the reduced set of feature values comprises:

performing a set of non-linear operations based on a kernel function to determine a kernel matrix;

determining a set of eigenvectors and eigenvalues based on the kernel matrix;

selecting a subset of eigenvectors and eigenvalues based on the set of eigenvectors and eigenvalues; and

updating the reduced set of feature values based on the subset of eigenvectors and eigenvalues.

10 . The method of claim 5 , further comprising:

obtaining the set of feature primitive stacks via a user input;

determining a resource consumption value based on the set of feature primitive stacks; and

selecting the first dataset of a plurality of datasets based on the resource consumption value.

11 . The method of claim 5 , further comprising:

receiving a third dataset;

determining whether the third dataset comprises the first feature set;

in response to a determination that the third dataset comprises the first feature set, retrieving a sequence of identifiers; and

generating a fourth dataset comprising the first feature set based on the sequence of identifiers.

12 . A set of non-transitory, machine-readable media storing program instructions that, when executed by a set of processors, causes the set of processors to perform operations comprising:

obtaining a first dataset comprising a first feature set;

generating a first set of feature values by providing the first dataset to a set of feature primitive stacks, wherein the set of feature primitive stacks does not comprise a second feature primitive stack;

determining a reduced set of feature values based on the first set of feature values by dimensionally reducing features of the first set of feature values;

generating an intermediate set of feature values by (i) providing a first value of the first dataset and a first value of the reduced set of feature values to at least one feature primitive of the set of feature primitive stacks and (ii) providing a second value of the first dataset and a second value of the reduced set of feature values to the second feature primitive stack;

updating the reduced set of feature values by dimensionally reducing features of the intermediate set of feature values; and

storing a second dataset comprising features of the intermediate set of feature values in association with the first feature set.

13 . The set of non-transitory, machine-readable media of claim 12 , wherein the set of feature primitive stacks is a first set of feature primitive stacks, the operations further comprising:

storing a first set of identifiers in a sequence of identifiers, wherein the first set of identifiers identify the first set of feature primitive stacks;

storing a second set of identifiers in the sequence of identifiers, wherein the second set of identifiers identify a first set of dimension-reducing operators used to determine the reduced set of feature values;

storing a third set of identifiers in the sequence of identifiers, wherein the third set of identifiers identify a second set of feature primitive stacks used to generate the intermediate set of feature values;

storing a fourth set of identifiers in the sequence of identifiers, wherein the fourth set of identifiers identify a second set of dimension-reducing operators used to dimensionally reduce features of the intermediate set of feature values; and

storing the sequence of identifiers in association with the first feature set.

14 . The set of non-transitory, machine-readable media of claim 13 , the operations further comprising:

receiving a third dataset;

in response to a determination that the third dataset comprises the first feature set, retrieving the sequence of identifiers; and

generating a fourth dataset comprising the first feature set based on the sequence of identifiers.

15 . The set of non-transitory, machine-readable media of claim 14 , the operations further comprising:

training a first machine learning model based on the second dataset;

determining a first output by providing the first machine learning model with a validation dataset;

training a second machine learning model based on the fourth dataset;

determining a second output by providing the second machine learning model with the validation dataset;

determining a difference value based on the first output and the second output; and

in response to a determination that the difference value satisfies an alert threshold, associating an alert indication with the third dataset.

16 . The set of non-transitory, machine-readable media of claim 12 , wherein providing the first dataset to the set of feature primitive stacks comprises:

determining whether a first feature of the first feature set was previously provided to a first feature primitive of the set of feature primitive stacks; and

in response to a determination that the first feature of the first feature set was previously provided to the first feature primitive, retrieving a stored set of feature values associated with the first feature primitive, wherein the first set of feature values comprises the stored set of feature values.

17 . The set of non-transitory, machine-readable media of claim 12 , wherein dimensionally reducing features of the first set of feature values comprises:

selecting a subset of the first feature set; and

extracting the reduced set of feature values based on the subset of the first feature set.

18 . The set of non-transitory, machine-readable media of claim 12 , the operations further comprising:

determining a dataset size of the first dataset;

determining a predicted resource consumption based on the dataset size; and

determining the set of feature primitive stacks based on the predicted resource consumption.

19 . The set of non-transitory, machine-readable media of claim 12 , the operations further comprising:

obtaining the set of feature primitive stacks via a user input;

determining a resource consumption value based on the set of feature primitive stacks; and

selecting the first dataset of a plurality of datasets based on the resource consumption value.

20 . The method of claim 5 , further comprising:

determining a dataset size of the first dataset;

determining a predicted resource consumption based on the dataset size; and

determining the set of feature primitive stacks based on the predicted resource consumption.