Interface for visualizing and improving model performance

1. A method comprising:

receiving, via a graphical user interface, data characterizing user input specifying a cost of error of a prediction, a model type requirement, and a computational resource allocation for model development;

determining, based on the received data and using a model, a prediction of model performance;

rendering, within the graphical user interface, a plot including a first axis, a second axis, and a region characterizing the determined prediction of model performance, the first axis including a characterization of a first performance metric and the second axis including a characterization of a second performance metric; and

rendering, within the graphical user interface, a graphical object prompting a user for user input for a course of action.

2. The method of claim 1 , wherein the model is trained with training data obtained from observations of users utilizing a model development platform.

3. The method of claim 2 , wherein the training data includes metadata of historical training data, historical selected model requirements, historical computational resource allocations, and/or historical resultant model performance.

4. The method of claim 3 , wherein the metadata of historical training data includes a number of rows of the historical training data, a number of columns of the historical training data, a number of observed values for each variable in the historical training data, a standard deviation of values of the historical training data, and/or a metric of data skew of the historical training data.

5. The method of claim 1 , further comprising:

determining a balance metric characterizing a relative proportion of observed classes within a dataset; and rendering, within the graphical user interface, a line or region indicative of a target accuracy based on the determined balance metric.

6. The method of claim 1 , further comprising:

receiving user input characterizing a relative cost of false negative and relative cost of false positive; and rendering a line indicative of constant cost weighted according to the received user input characterizing the relative cost of false negative and relative cost of false positive.

7. The method of claim 1 , further comprising:

receiving user input characterizing a target accuracy;

determining a probability of developing a model according to the target accuracy, the model type requirement, and the computational resource allocation; and rendering, within the graphical user interface, a visualization characterizing the probability.

8. The method of claim 1 , further comprising:

determining, by a model generator, a candidate model based on the received data characterizing the user input and a dataset;

determining a performance of the candidate model for each of a plurality of subgroups, the determining based on the dataset; and

identifying subgroups for which the candidate model underperforms relative to the plurality of subgroups or for which the candidate model over performs relative to the plurality of subgroups.

9. The method of claim 8 , further comprising:

rendering, within the graphical user interface, a visualization of relative performance of the identified subgroups; and

wherein the course of action includes terminating further training of the model, creating a split-model, and/or removing the subgroup from the dataset.

10. The method of claim 1 , further comprising:

rendering a first line indicative of target accuracy, a second line indicative of constant accuracy, and a third line indicative of constant cost; and receiving data characterizing a target accuracy and a performance metric of a candidate model; rendering a region indicative of the target accuracy, the region indicative of the target accuracy bounded by at least:

the first line indicative of the target accuracy and an origin of the plot;

a second line indicative of constant accuracy and the origin; or

a second line indicative of constant accuracy, the third line indicative of constant cost, and the origin;

wherein the first performance metric includes rate of false positive, count of false positive, cost of false positive, benefit missed by false positive, true positive, benefit of true positive, benefit of minimizing false positive, or benefit of maximizing true positive;

wherein the second performance metric includes rate of false negative, count of false negative, cost of false negative, benefit missed by false negative, true negative, benefit of true negative, benefit of minimizing false negative, or benefit of maximizing true negative.

11. The method of claim 1 , further comprising:

rendering, within the plot, a plurality of lines each indicative of a respective constant cost or a respective constant accuracy.

12. A system comprising:

at least one processor;

memory storing instructions which, when executed by the at least one processor, causes the at least one processor to perform operations comprising:

receiving, via a graphical user interface, data characterizing user input specifying a cost of error of a prediction, a model type requirement, and a computational resource allocation for model development;

determining, based on the received data and using a model, a prediction of model performance;

rendering, within the graphical user interface, a plot including a first axis, a second axis, and a region characterizing the determined prediction of model performance, the first axis including a characterization of a first performance metric and the second axis including a characterization of a second performance metric; and

rendering, within the graphical user interface, a graphical object prompting a user for user input for a course of action.

13. The system of claim 12 , wherein the model is trained with training data obtained from observations of users utilizing a model development platform.

14. The system of claim 13 , wherein the training data includes metadata of historical training data, historical selected model requirements, historical computational resource allocations, and/or historical resultant model performance.

15. The system of claim 14 , wherein the metadata of historical training data includes a number of rows of the historical training data, a number of columns of the historical training data, a number of observed values for each variable in the historical training data, a standard deviation of values of the historical training data, and/or a metric of data skew of the historical training data.

16. The system of claim 12 , the operations further comprising:

determining a balance metric characterizing a relative proportion of observed classes within a dataset; and rendering, within the graphical user interface, a line or region indicative of a target accuracy based on the determined balance metric.

17. The system of claim 12 , the operations further comprising:

receiving user input characterizing a relative cost of false negative and relative cost of false positive; and rendering a line indicative of constant cost weighted according to the received user input characterizing the relative cost of false negative and relative cost of false positive.

18. The system of claim 12 , the operations further comprising:

determining, by a model generator, a candidate model based on the received data characterizing the user input and a dataset;

determining a performance of the candidate model for each of a plurality of subgroups, the determining based on the dataset; and

identifying subgroups for which the candidate model underperforms relative to the plurality of subgroups or for which the candidate model over performs relative to the plurality of subgroups.

19. The system of claim 18 , the operations further comprising:

rendering, within the graphical user interface, a visualization of relative performance of the identified subgroups; and

wherein the course of action includes terminating further training of the model, creating a split-model, and/or removing the subgroup from the dataset.

20. A non-transitory computer program product storing instructions, which when executed by at least one processor of at least one computing system, cause the at least one processor to perform operations comprising:

receiving, via a graphical user interface, data characterizing user input specifying a cost of error of a prediction, a model type requirement, and a computational resource allocation for model development;

determining, based on the received data and using a model, a prediction of model performance;

rendering, within the graphical user interface, a plot including a first axis, a second axis, and a region characterizing the determined prediction of model performance, the first axis including a characterization of a first performance metric and the second axis including a characterization of a second performance metric; and

rendering, within the graphical user interface, a graphical object prompting a user for user input for a course of action.