Automated machine learning feature processing

A non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a machine to perform operations comprising:

• selecting a predefined number of machine-based transactions;
• determining a first set of features based on the predefined number of machine-based transactions;
• applying one or more machine learning feature selection algorithms to the first set of features to determine a second set of features, wherein the second set of features is smaller in number than the first set of features, and the second set of features includes no more than a predefined number of features, the predefined number of features is smaller in number than the first set of features;
• constructing a decision tree based on the second set of features, wherein the decision tree has a depth that is less than a predefined depth, and the decision tree includes a plurality of nodes corresponding to the second set of features;
• applying a test data set to the decision tree to determine an accuracy level for each node included in the decision tree, wherein the accuracy level for each node indicates whether that node correctly identifies data in the test data set being associated with a fraudulent or a non-fraudulent machine-based transaction;
• selecting, from the plurality of nodes, a subset of nodes with each node in the subset of nodes having the accuracy level greater than a predefined accuracy level;
• determining a third set of features associated with the subset of nodes;
• and
• generating a set of the human-readable transaction categorization rules based on the third set of features, wherein a total number of the human-readable transaction categorization rules included in the set is fewer than a limit on a total number of human readable transaction categorization rules to be generated.

A system, comprising:

• a non-transitory memory;
• and
• one or more hardware processors coupled to the non-transitory memory and configured to execute instructions to perform operations comprising:
• selecting a predefined number of machine-based transactions;
• determining a first set of features based on the predefined number of machine-based transactions;
• determining, using one or more machine learning feature selection algorithms, a second set of features from the first set of features, wherein the second set of features is smaller in number than the first set of features, the second set of features includes no more than a predefined number of features, and the predefined number of features is smaller than a number of the first set of features;
• constructing a decision tree based on the second set of features, wherein the decision tree has a depth that is less than a predefined depth and the decision tree includes a plurality of nodes corresponding to the second set of features;
• applying a test data set to the decision tree to determine an accuracy level for each node included in the decision tree;
• selecting, from the plurality of nodes, a subset of nodes with each node in the subset of nodes having the accuracy level greater than a predefined accuracy level, wherein the accuracy level for each node in the subset of nodes indicates whether that node correctly identifies data in the test data set being associated with a fraudulent or a non-fraudulent machine-based transaction;
• determining a third set of features associated with the subset of nodes;
• and
• generating a set of the human-readable transaction categorization rules based on the third set of features, wherein a total number of the human-readable transaction categorization rules included in the set is fewer than a limit on a total number of human readable transaction categorization rules to be generated.