System and method for generating an effective test data set for testing big data applications
A system for generating an effective test data set for testing big data applications includes a data collector, a data analyzer, an input domain modeler, a self-adaptive input domain modeler, and a test data set generator. The data collector collects a high volume of data from an original data set and initial constraints, the data analyzer analyzes the data and the initial constraints to generate analytical results, the input domain modeler automatically generates an input domain model based on the analytical results, the self-adaptive input domain modeler generates a self-adaptive input domain model by combining the input domain model and analytical results, and the test data set generator generates an initial test data set based on the self-adaptive input domain model. A method for generating an effective test data set for testing big data applications is also described.
1. A system comprising:
a data collector for collecting initial constraints and a high volume of data from an original data set, wherein the initial constraints comprise one or more of statistical-based constraints, foreign key constraints, logic constraints, and density constraints;
a data analyzer, including a processor, for analyzing the high volume of data and the initial constraints to generate analytical results;
an input domain modeler for generating an input domain model by finding at least one input domain, dividing each input domain into blocks, and applying combinatorial coverage criteria to the blocks;
a self-adaptive input domain modeler for generating a self-adaptive input domain model by combining the input domain model and the analytical results; and
a test data set generator for generating an initial test data set based on the self-adaptive input domain model,
wherein:
the initial test data set is reviewed to determine how effective the test data set is, the effectiveness of the initial test data set being measured based on the number of initial constraints satisfied by the test data set and the strength of the satisfied initial constraints;
if the initial test data set is not effective enough, the self-adaptive input domain modeler applies a machine-learning technique to adjust the self-adaptive input domain model based on feedback from at least one reviewer and updates the self-adaptive input domain model based on new data or constraints; and
the test data set generator generates a subsequent test data set based on the updated self-adaptive input domain model.
2. The system of claim 1 , wherein the analytical results comprise relationships among objects and a statistical distribution of the data.
3. The system of claim 2 , wherein the input domain model is based on the object relationships and statistical distribution.
4. The system of claim 1 , wherein if the subsequent test data set is not effective enough, the self-adaptive input domain modeler applies a machine-learning technique to adjust the self-adaptive input domain model based on feedback from at least one reviewer.
5. The system of claim 1 , wherein the initial constraints are specified in a JSON format.
6. A computer-implemented method comprising:
collecting initial constraints and a high volume of data from an original data set, wherein the initial constraints comprise one or more of statistical-based constraints, foreign key constraints, logic constraints, and density constraints;
analyzing, using a processor, the high volume of data and the initial constraints to generate analytical results;
generating an input domain model by finding at least one input domain, dividing each input domain into blocks, and applying combinatorial coverage criteria to the blocks;
generating a self-adaptive input domain model by combining the input domain model and the analytical results;
generating an initial test data set based on the self-adaptive input domain model;
reviewing the initial effective test data set to determine how effective the test data set is, wherein the effectiveness of the initial test data set is measured based on the number of initial constraints satisfied by the test data set and the strength of the satisfied initial constraints; and
if the initial test data set is not effective enough,
applying a machine-learning technique to adjust the self-adaptive input domain model based on feedback from at least one reviewer;
updating the self-adaptive input domain model based on new data or constraints; and
generating a subsequent test data set based on the updated self-adaptive input domain model.
7. The method of claim 6 , wherein the analytical results comprise relationships among objects and a statistical distribution of the data.
8. The method of claim 7 , wherein the input domain model is based on the object relationships and statistical distribution.
9. The method of claim 6 , wherein if the subsequent test data set is not effective enough, applying a machine-learning technique to adjust the self-adaptive input domain model based on feedback from at least one reviewer.
10. The method of claim 6 , wherein the initial constraints are specified in a JSON format.