Computer system to generate scalable plots using clustering

A computer-implemented method, comprising:

• receiving an indication to generate Individual Conditional Expectation (ICE) plots, the indication received based on an interaction with a graphical user interface (GUI);
• identifying a dataset and a model to generate the ICE plots, the dataset comprising observations and the ICE plots to visually indicate an effect that a variable has on a predicted outcome;
• identifying a range of values for the variable to compute individual curves for the observations;
• computing individual curves for the observations of the dataset, wherein an individual curve is computed for each observation by varying the variable over the range of values for the observation using the model;
• performing segmenting of the individual curves to generate a number of clusters of curves, each cluster of curves comprising a subset of the individual curves and each of the subsets of the individual curves represented by a respective proxy curve;
• plotting each of the proxy curves to visually indicate the effect the variable has on the predicted outcome;
• presenting the ICE plots of the proxy curves in the GUI of a display on a display device, and each respective proxy curve is selectable to drill down on one of the subsets of individual curves within the cluster of curves;
• receiving a selection of a subset of individual curves via another interaction with the GUI;
• and
• presenting the subset of individual curves selected in the GUI.

At least one non-transitory computer-readable storage medium comprising instructions that when executed cause processing circuitry to:

• receive an indication to generate Individual Conditional Expectation (ICE) plots, the indication received based on an interaction with a graphical user interface (GUI);
• identify a dataset and a model to generate the ICE plots, the dataset comprising observations and the ICE plots to visually indicate an effect that a variable has on a predicted outcome;
• identify a range of values for the variable to compute individual curves for the observations;
• compute individual curves for the observations of the dataset, wherein an individual curve is computed for each observation by varying the variable over the range of values for the observation using the model;
• perform segmenting of the individual curves to generate a number of clusters of curves, each cluster of curves comprising a subset of the individual curves and each of the subsets of the individual curves represented by a respective proxy curve;
• plot each of the proxy curves to visually indicate the effect the variable has on the predicted outcome;
• present the ICE plots of the proxy curves in the GUI of a display on a display device, and each respective proxy curve is selectable to drill down on one of the subsets of individual curves within the cluster of curves;
• receive a selection of a subset of individual curves via another interaction with the GUI;
• and
• present the subset of individual curves selected in the GUI.