MACHINE-LEARNING TECHNIQUES FOR PREDICTING ACTIONS AND BEHAVIOR
Techniques are disclosed for specifying a predictive goal that describes an action that a customer of an organization may take or a behavior that a customer of the organization may exhibit. A training engine receives the predictive goal from a user via a Graphical User Interface (GUI). The user specifies the predictive goal by entering one or more customizable logical or comparative statements. A training engine divides historical data into training data and testing data, selects useful and relevant features from the historical data as inputs to a machine learning model, and trains the machine learning model. The training engine transmits the trained machine learning model to an inference engine that generates, for each customer in a historical data set, a probability that the customer will satisfy the predictive goal within a specified time frame. The disclosed techniques further generate explanatory data associated with the trained machine learning model.
1 . A computer-implemented method for predicting actions and behaviors, the method comprising:
receiving, from a user via a graphical user interface, a plurality of prediction goal criteria for a machine learning model;
generating, based on the plurality of prediction goal criteria, a prediction goal;
generating a plurality of machine learning features based on data included in a historical data set;
generating, based on the plurality of machine learning features and the historical data set, a trained machine learning model; and
predicting, via the trained machine learning model, a probability that the prediction goal will be satisfied within a specified time frame.
2 . The computer-implemented method of claim 1 , wherein each of the plurality of prediction goal criteria includes at least one of an event or a property to be predicted.
3 . The computer-implemented method of claim 1 , wherein two or more of the plurality of prediction goal criteria are included as operands in a logical operation.
4 . The computer-implemented method of claim 1 , further comprising:
partitioning, via a training engine, the historical data set into data associated with a training time period and data associated with a testing time period;
determining, via the training engine, that the training time period and the testing time period do not overlap in time;
determining, via the training engine and based on the specified time frame, that the data associated with the testing time period is not right-censored;
determining, via the training engine and based on the specified time frame, that a duration associated with the training time period is longer than a first predetermined threshold; and
determining, via the training engine and based on the specified time frame, that a duration associated with the testing time period is longer than a second predetermined threshold.
5 . The computer-implemented method of claim 1 , wherein generating the plurality of machine learning features further comprises:
calculating, via a training engine, a predictive strength for each of one or more machine learning features; and
selecting, via the training engine, at least one of the one or more machine learning features for inclusion in the plurality of machine learning features based on the predictive strength.
6 . The computer-implemented method of claim 1 , wherein generating the plurality of machine learning features further comprises:
calculating, via a training engine, a correlation value between two or more machine learning features; and
selecting, via the training engine, at least one of the one or more machine learning features for inclusion in the plurality of machine learning features based on the correlation value.
7 . The computer-implemented method of claim 1 , wherein the historical data set includes at least one of customer data, transaction data, and organizational data.
8 . One or more non-transitory computer-readable media storing instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of:
receiving, from a user via a graphical user interface, a plurality of prediction goal criteria for a machine learning model;
generating, based on the plurality of prediction goal criteria, a prediction goal;
generating a plurality of machine learning features based on data included in a historical data set;
generating, based on the plurality of machine learning features and the historical data set, a trained machine learning model; and
predicting, via the trained machine learning model, a probability that the prediction goal will be satisfied within a specified time frame.
9 . The one or more non-transitory computer-readable media of claim 8 , wherein each of the plurality of prediction goal criteria includes at least one of an event or a property to be predicted.
10 . The one or more non-transitory computer-readable media of claim 8 , wherein two or more of the plurality of prediction goal criteria are included as operands in a logical operation.
11 . The one or more non-transitory computer-readable media of claim 8 , wherein the instructions further cause the one or more processors to perform the steps of:
partitioning, via a training engine, the historical data set into data associated with a training time period and data associated with a testing time period;
determining, via the training engine, that the training time period and the testing time period do not overlap in time;
determining, via the training engine and based on the specified time frame, that the data associated with the testing time period is not right-censored;
determining, via the training engine and based on the specified time frame, that a duration associated with the training time period is longer than a first predetermined threshold; and
determining, via the training engine and based on the specified time frame, that a duration associated with the testing time period is longer than a second predetermined threshold.
12 . The one or more non-transitory computer-readable media of claim 8 , wherein the instructions, when generating the plurality of machine learning features, further cause the one or more processors to perform the steps of:
calculating, via a training engine, a predictive strength for each of one or more machine learning features; and
selecting, via the training engine, at least one of the one or more machine learning features for inclusion in the plurality of machine learning features based on the predictive strength.
13 . The one or more non-transitory computer-readable media of claim 8 , wherein the instructions, when generating the plurality of machine learning features, further cause the one or more processors to perform the steps of:
calculating, via a training engine, a correlation value between two or more machine learning features; and
selecting, via the training engine, at least one of the one or more machine learning features for inclusion in the plurality of machine learning features based on the correlation value.
14 . The one or more non-transitory computer-readable media of claim 8 , wherein the historical data set includes at least one of customer data, transaction data, and organizational data.
15 . A system comprising:
one or more memories storing instructions; and
one or more processors for executing the instructions to:
receive, from a user via a graphical user interface, a plurality of prediction goal criteria for a machine learning model;
generate, based on the plurality of prediction goal criteria, a prediction goal;
generate a plurality of machine learning features based on data included in a historical data set;
generate, based on the plurality of machine learning features and the historical data set, a trained machine learning model; and
predict, via the trained machine learning model, a probability that the prediction goal will be satisfied within a specified time frame.
16 . The system of claim 15 , wherein each of the plurality of prediction goal criteria includes at least one of an event or a property to be predicted.
17 . The system of claim 15 , wherein two or more of the plurality of prediction goal criteria are included as operands in a logical operation.
18 . The system of claim 15 , wherein the one or more processors further execute the instructions to:
partition, via a training engine, the historical data set into data associated with a training time period and data associated with a testing time period;
determine, via the training engine, that the training time period and the testing time period do not overlap in time;
determine, via the training engine and based on the specified time frame, that the data associated with the testing time period is not right-censored;
determine, via the training engine and based on the specified time frame, that a duration associated with the training time period is longer than a first predetermined threshold; and
determine, via the training engine and based on the specified time frame, that a duration associated with the testing time period is longer than a second predetermined threshold.
19 . The system of claim 15 , wherein the one or more processors, when generating the plurality of machine learning features, further execute the instructions to:
calculate, via a training engine, a predictive strength for each of one or more machine learning features; and
select, via the training engine, at least one of the one or more machine learning features for inclusion in the plurality of machine learning features based on the predictive strength.
20 . The system of claim 15 , wherein the one or more processors, when generating the plurality of machine learning features, further execute the instructions to:
calculate, via a training engine, a correlation value between two or more machine learning features; and
select, via the training engine, at least one of the one or more machine learning features for inclusion in the plurality of machine learning features based on the correlation value.