Generating recommended inputs
Example embodiments relate to generating sets of recommended inputs for changing predicted results of a predictive model. The examples disclosed herein access, from a database, a historical set of inputs and results of a predictive model. A function is approximated based on the historical set of inputs and results, and a gradient of the function is computed using a result of the function with respect to a local maximum value of the function. A set of recommended inputs is generated based on the gradient of the function, where a recommended input produces a positive result of the function.
1. A method for execution by a computing device for generating recommended inputs, comprising:
accessing, from a database, a historical set of inputs and results of a predictive model;
approximating a function based on the historical set of inputs and results;
computing a gradient of the function using a result of the function with respect to local maximum values of the function;
generating a set of recommended inputs based on the gradient of the function,
wherein using a recommended input in the function produces a positive result of the function; and
generating a visualization of a graph of the function,
wherein the visualization indicates the local maximum values of the function, the set of recommended inputs and the gradient provided between the local maximum values of the function,
wherein the gradient represents a slope of a tangent of the graph of the function, and
wherein the visualization includes a non-zero threshold result value intersecting the graph of the function and the recommended inputs have values greater than or equal to the non-zero threshold result value.
2. The method of claim 1 , wherein the predictive model is created from a set of labeled cases, wherein each case comprises a set of attributes and corresponding results.
3. The method of claim 1 , wherein the historical set of inputs and results comprises positive and negative values.
4. The method of claim 1 , further comprising enriching the historical set of inputs and results by multiplying each result of the historical set of inputs and results with a corresponding confidence value.
5. The method of claim 4 , wherein the set of recommended inputs is prioritized based on the corresponding confidence values of the result of the predictive model for each recommended input.
6. The method of claim 4 , further comprising reducing the function to three dimensions using two modifiable inputs of the historical set of inputs.
7. The method of claim 6 , wherein reducing the function to three dimensions includes utilizing feature selection methods or feature extraction methods.
8. The method of claim 1 , further comprising performing Multivariate Lagrange interpolation on the historical set of inputs and results to approximate the function.
9. The method of claim 1 , wherein gradient points of the gradient of the function are in a direction of a greatest rate of increase of the function.
10. The method of claim 1 , wherein the gradient indicates a direction of each of the local maximum values of the function.
11. A non-transitory machine-readable storage medium encoded with instructions executable by a processor of a computing device, the non-transitory machine-readable storage medium comprising instructions to:
access, from a database, a historical set of inputs and positive and negative results, and a corresponding confidence values for each result of a predictive model,
wherein the predictive model is created from a set of labeled cases, and
wherein each case comprises a set of attributes and corresponding results;
enrich the historical set of inputs and positive and negative results by multiplying each result of the historical set of inputs and positive and negative results with the corresponding confidence value;
approximate a function based on the enriched historical set of inputs and positive and negative results;
compute a gradient of the function using a result of the function with respect to local maximum values of the function;
generate a set of recommended inputs based on the gradient of the function,
wherein using a recommended input in the function produces a positive result of the function, and
generate a visualization of a graph of the function,
wherein the visualization indicates the local maximum values of the function, the set of recommended inputs and the gradient provided between the local maximum values of the function,
wherein the gradient represents a slope of a tangent of the graph of the function, and
wherein the visualization includes a non-zero threshold result value intersecting the graph of the function and the recommended inputs have values greater than or equal to the non-zero threshold result value.
12. The non-transitory machine-readable storage medium of claim 11 , further comprising instructions to reduce the function to three dimensions using two modifiable inputs of the historical set of inputs.
13. The non-transitory machine-readable storage medium of claim 12 , wherein the instructions to reduce the function to three dimensions include utilizing feature selection methods or feature extraction methods.
14. The non-transitory machine-readable storage medium of claim 11 , further comprising instructions to perform Multivariate Lagrange interpolation on the historical set of inputs and results to approximate the function.
15. The non-transitory machine-readable storage medium of claim 11 , wherein gradient points of the gradient of the function are in a direction of a greatest rate of increase of the function.
16. A computing device for generating recommended inputs, the computing device comprising:
a historical results engine to access, from a database, a historical set of inputs and positive and negative results, and a corresponding confidence value for each result of a predictive model and to enrich the historical inputs and positive and negative results by multiplying each historical positive and negative result with the corresponding confidence value;
a function engine to perform Multivariate LaGrange interpolation on the historical set of inputs and positive and negative results to approximate a function;
a gradient engine to compute a gradient of the function using a result of the function with respect to local maximum values of the function; and
an input engine to generate a set of recommended inputs based on the gradient of the function,
wherein using a recommended input value in the function produces a positive result of the function,
wherein the gradient engine generates a visualization of a graph of the function,
wherein the visualization indicates the local maximum values of the function, the set of recommended inputs and the gradient provided between the local maximum values of the function,
wherein the gradient represents a slope of a tangent of the graph of the function, and
wherein the visualization includes a non-zero threshold result value intersecting the graph of the function and the recommended inputs have values greater than or equal to the non-zero threshold result value.
17. The computing device of claim 16 , wherein the function engine is to reduce the function to three dimensions using two modifiable inputs of the historical set of inputs.
18. The computing device of claim 17 , wherein the function engine reduces the function to three dimensions utilizing feature selection methods or feature extraction methods.
19. The computing device of claim 16 , wherein the set of recommended inputs is prioritized based on the corresponding confidence values of the result of the predictive model for each recommended input.
20. The computing device of claim 16 , wherein gradient points of the gradient of the function are in a direction of a greatest rate of increase in the function.