Advanced training of machine-learning models usable in control systems and other systems

A method comprising:

• receiving, by a processor, a first training dataset that includes (i) independent-variable values corresponding to independent variables and (ii) dependent-variable values corresponding to a dependent variable that is influenced by the independent variables, wherein the independent-variable values include linear-variable values corresponding to at least one linear independent variable that has a linear relationship with the dependent variable, and wherein the independent-variable values include nonlinear-variable values corresponding to at least one nonlinear independent variable that has a nonlinear relationship with the dependent variable;
• determining, by the processor, cluster assignments for the nonlinear-variable values by applying a clustering algorithm to the nonlinear-variable values, each nonlinear-variable value among the nonlinear-variable values being assigned to a respective cluster among a plurality of clusters using the clustering algorithm;
• generating, by the processor, a second training dataset indicating the dependent-variable values, the linear-variable values, and the cluster assignments for the nonlinear-variable values;
• training, by the processor, a model based on the second training dataset by tuning one or more weights of the model;
• subsequent to training the model:
• receiving, by the processor, a first data input excluding a dependent-variable value corresponding to the dependent variable and including a set of independent-variable values corresponding to the independent variables, wherein the set of independent-variable values includes at least one linear-variable value corresponding to the at least one linear independent variable and at least one nonlinear-variable value corresponding to the at least one nonlinear independent variable;
• determining, by the processor, at least one cluster assignment for the at least one nonlinear-variable value, from among the plurality of clusters determined using the clustering algorithm;
• generating, by the processor, a second data input indicating the at least one linear-variable value and the at least one cluster assignment for the at least one nonlinear-variable value;
• applying, by the processor, the trained model to the second data input to predict the dependent-variable value associated with the second data input;
• and
• transmitting, by the processor, an electronic signal to an electronic device, the electronic signal indicating the dependent-variable value predicted via the trained model.

A non-transitory computer-readable medium comprising program code that is executable by a processor for causing the processor to:

• receive a first training dataset that includes (i) independent-variable values corresponding to independent variables and (ii) dependent-variable values corresponding to a dependent variable that is influenced by the independent variables, wherein the independent-variable values include linear-variable values corresponding to at least one linear independent variable that has a linear relationship with the dependent variable, and wherein the independent-variable values include nonlinear-variable values corresponding to at least one nonlinear independent variable that has a nonlinear relationship with the dependent variable;
• determine cluster assignments for the nonlinear-variable values by applying a clustering algorithm to the nonlinear-variable values, each nonlinear-variable value among the nonlinear-variable values being assigned to a respective cluster among a plurality of clusters using the clustering algorithm;
• generate a second training dataset indicating the dependent-variable values, the linear-variable values, and the cluster assignments for the nonlinear-variable values;
• train a model based on the second training dataset by tuning one or more weights of the model;
• subsequent to training the model:
• receive a first data input excluding a dependent-variable value corresponding to the dependent variable and including a set of independent-variable values corresponding to the independent variables, wherein the set of independent-variable values includes at least one linear-variable value corresponding to the at least one linear independent variable and at least one nonlinear-variable value corresponding to the at least one nonlinear independent variable;
• determine at least one cluster assignment for the at least one nonlinear-variable value, from among the plurality of clusters determined using the clustering algorithm;
• generate a second data input indicating the at least one linear-variable value and the at least one cluster assignment for the at least one nonlinear-variable value;
• apply the trained model to the second data input to predict the dependent-variable value associated with the second data input;
• and
• transmit an electronic signal to an electronic device, the electronic signal indicating the dependent-variable value predicted via the trained model.