System and method for auto-query generation
Various systems and methods provide an intuitive user interface that enables automatic specification of queries and constraints for analysis by ML component. Various implementations provide methodologies for automatically formulating machine learning (“ML”) and optimization queries. The automatic generation of ML and/or optimization queries can be configured to use examples to facilitate formulation of ML and optimization queries. One example method includes accepting input data specifying variables and data values associated with the variables. Within the input data any unspecified data records are identified, and a relationship between the variables specified in the input data and a variable associated with the at least one unspecified data record is automatically determined. The relationship can be automatically determined based on training data contained within the input data. Once a relationship is established a ML problem can be automatically generated.
1. A computer implemented method, comprising:
accessing, by a computer system, input data comprising:
training data comprising a first plurality of data records, each of the first plurality of data records having a numeric data value for each of a plurality of variables; and
a second plurality of data records, each of the second plurality of data records having a numeric data value for at least one of the plurality of variables;
identifying, by the computer system, at least a first not fully specified data record within the second plurality of data records that does not include at least a first numeric data value for at least a first variable of the plurality of variables;
determining, automatically by the computer system and based on the training data, a relationship between the first variable and at least one other variable in the plurality of variables, wherein determining the relationship comprises:
generating, using a machine learning technique and the first plurality of records in the training data, a mathematical model representing the relationship between the first variable and the at least one other variable; and
obtaining, using the mathematical model representing the relationship between the first variable and the at least one other variable, at least the first numeric data value for the first variable in the first not fully specified data record.
2. A system comprising:
at least one processor configured to perform:
accessing input data comprising:
training data comprising a first plurality of data records, each of the first plurality of data records having a numeric data value for each of a plurality of variables; and
a second plurality of data records, each of the second plurality of data records having a numeric data value for at least one of the plurality of variables;
identifying at least a first not fully specified data record within the second plurality of data records that does not include at least a first numeric data value for at least a first variable of the plurality of variables;
determining, automatically and based on the training data, a relationship between the first variable and at least one other variable in the plurality of variables, wherein determining the relationship comprises:
generating, using a machine learning technique and the first plurality of records in the training data, a mathematical model representing the relationship between the first variable and the at least one other variable; and
obtaining, using the mathematical model representing the relationship between the first variable and the at least one other variable, at least the first numeric data value for the first variable in the first not fully specified data record.
3. The method according to claim 1 , wherein generating the mathematical model comprises identifying a function for calculating a value of the first variable given at least one value for the at least one other variable.
4. The method according to claim 1 , further comprising displaying, by the computer system, the relationship between the first variable and the at least one other variable.
5. The method according to claim 1 , wherein accessing the input data includes receiving, within a spreadsheet user interface, a selection of the first plurality of data records and the second plurality of data records.
6. The method according to claim 1 , further comprising determining a confidence value associated with the first numeric data value.
7. The method according to claim 1 , wherein generating the mathematical model representing the relationship between the first variable and the at least one other variable is performed by using symbolic regression.
8. The method according to claim 1 , wherein accessing the input data is performed using a spreadsheet user interface, and wherein generating the mathematical model includes:
generating one or more candidate query patterns that relate a blank cell in the spreadsheet user interface to its surrounding non-blank cells in the spreadsheet user interface;
for each pattern in the one or more candidate query patterns:
collecting data from the spreadsheet user interface that matches the pattern,
generating, using the machine learning technique and the collected data, a respective candidate mathematical model representing a relationship between a variable associated with the blank cell and one or more variables associated with the non-blank cells;
determining a confidence associated with the generated candidate mathematical model;
and
selecting the generated candidate mathematical model having a highest confidence level as the mathematical model.
9. The method according to claim 1 , wherein the first not fully specified data record does not include a second numeric data value for a second variable of the plurality of variables, the method further comprising:
determining the second numeric data value based, at least in part, on the first numeric data value.
10. The method according to claim 1 , wherein generating the mathematical model representing the relationship between the first variable and the at least one other variable comprises generating a mathematical model that may be used for estimating a value of the first variable given at least one respective value for the at least one other variable.
11. The method according to claim 1 ,
wherein the first not fully specified data record does not include a second numeric data value for a second variable of the plurality of variables and does include a third numeric data value for a third variable of the plurality of variables, and wherein the obtaining comprises obtaining, using the mathematical model and at least one optimization technique, the first numeric data value and the second numeric data value based on the third numeric data value.
12. The system according to claim 2 , wherein generating the mathematical model comprises identifying a function for calculating a value of the first variable given at least one value for the at least one other variable.
13. The system according to claim 2 , further comprising a display, wherein the at least one processor is configured to cause the display to show the relationship between the first variable and the at least one other variable.
14. The system according to claim 2 , wherein accessing the input data includes receiving, within a spreadsheet user interface, a selection of the first plurality of data records and the second plurality of data records.
15. The system according to claim 2 , wherein the at least one processor is further configured to determine a confidence value associated with the first numeric data value.
16. The system according to claim 2 , wherein generating the mathematical model representing the relationship between the first variable and the at least one other variable is performed by using symbolic regression.
17. The system according to claim 2 , wherein accessing the input data is performed using a spreadsheet user interface, and wherein generating the mathematical model includes:
generating one or more candidate query patterns that relate a blank cell in the spreadsheet user interface to its surrounding non-blank cells in the spreadsheet user interface;
for each pattern in the one or more candidate query patterns:
collecting data from the spreadsheet user interface that matches the pattern,
generating, using the machine learning technique and the collected data, a respective candidate mathematical model representing a relationship between a variable associated with the blank cell and one or more variables associated with the non-blank cells;
determining a confidence associated with the generated candidate mathematical model;
and
selecting the generated candidate mathematical model having a highest confidence level as the mathematical model.
18. The system according to claim 2 , wherein the first not fully specified data record does not include a second numeric data value for a second variable of the plurality of variables, and wherein the at least one processor is further configured to perform determining the second numeric data value based, at least in part, on the first numeric data value.
19. The system according to claim 2 , wherein generating the mathematical model representing the relationship between the first variable and the at least one other variable comprises generating a mathematical model that may be used for estimating a value of the first variable given at least one respective value for the at least one other variable.
20. The system according to claim 2 ,
wherein the first not fully specified data record does not include a second numeric data value for a second variable of the plurality of variables and does include a third numeric data value for a third variable of the plurality of variables, and
wherein the obtaining comprises obtaining, using the mathematical model and at least one optimization technique, the first numeric data value and the second numeric data value based on the third numeric data value.
21. The method according to claim 5 , wherein the identifying, determining, and obtaining occur in response to receiving the selection.
22. The method according to claim 6 , further comprising encoding, visually, the confidence value associated with the first numeric data value.
23. The method according to claim 8 , wherein generating the one or more candidate query patterns includes generating one or more patterns including both absolute cell positions and cell positions that are relative to the first not fully specified data record in the spreadsheet user interface.
24. The method according to claim 10 , comprising using the mathematical model to estimate values of the first variable in the second plurality data records based on at least one value of the at least one other variable in the second plurality of data records.
25. The method according to claim 11 , wherein the optimization technique comprises a gradient search technique.
26. The method according to claim 11 , wherein obtaining the first and second numeric data values is performed at least in part by using the mathematical model to compute an estimate of the third numeric data value based, at least in part, on the first and second numeric data values and comparing the estimate of the third numeric data value to the third numeric data value.
27. The method according to claim 11 , wherein the first not fully specified data record includes a fourth numeric data value for a fourth variable of the plurality of variables, and wherein obtaining the first and second numeric data values comprises using the fourth numeric data value as a search constraint in the optimization technique.
28. The system according to claim 14 , wherein the identifying, determining, and obtaining occur in response to receiving the selection.
29. The system according to claim 15 , wherein the at least one processor is further configured to encode, visually, the confidence value associated with the first numeric data value.
30. The system according to claim 17 , wherein generating the one or more candidate query patterns includes generating one or more patterns including both absolute cell positions and cell positions that are relative to the first not fully specified at least one unspecified data record in the spreadsheet user interface.
31. The system according to claim 19 , wherein the at least one processor is configured to perform: using the mathematical model to estimate values of the first variable in the second plurality data records based on at least one value of the at least one other variable in the second plurality of data records.
32. The system according to claim 20 , wherein the optimization technique comprises a gradient search technique.
33. The system according to claim 20 , wherein obtaining the first and second numeric data values is performed at least in part by using the mathematical model to compute an estimate of the third numeric data value based, at least in part, on the first and second numeric data values and comparing the estimate of the third numeric data value to the third numeric data value.
34. The system according to claim 20 , wherein the first not fully specified data record includes a fourth numeric data value for a fourth variable of the plurality of variables, and wherein obtaining the first and second numeric data values comprises using the fourth numeric data value as a search constraint in the optimization technique.
35. The method according to claim 22 , wherein encoding visually the confidence value associated with the first numeric data value includes displaying the first numeric data value with an intensity that depends on the confidence value.
36. The system according to claim 29 , wherein encoding visually the confidence value associated with the first numeric data value includes displaying the first numeric data value with an intensity that depends on the confidence value.