Machine learning to emulate software applications
Methods and systems for emulating an application include generating a log template to match one or more patterns in a set of application logs collected from an original application. Semantic state representations are learned for the original application from the log template. A classifier is trained to predict a next action template based on a sequence of prior action templates. A regressor is trained to generate a parameter value for a template based on a sequence of prior action templates and a particular semantic state of the original application.
1. A computer-implemented method for emulating an application, comprising:
generating a log template to match one or more patterns in a set of application logs collected from an original application;
learning semantic state representations for the original application from the log template;
training a classifier to predict a next action template based on a sequence of prior action templates;
training a regressor to generate a parameter value for a template based on a sequence of prior action templates and a particular semantic state of the original application; and
deploying the classifier, the regressor, and the log template to a remote computer system for execution to emulate the original application.
2. The computer-implemented method of claim 1 , wherein the classifier is a recurrent neural network model.
3. The computer-implemented method of claim 2 , wherein the recurrent neural network model is implemented as a long-short term memory (LSTM) model or a transformer model.
4. The computer-implemented method of claim 1 , further comprising collecting the set of application logs from the original application during operation of the original application.
5. The computer-implemented method of claim 4 , wherein the set of application logs include requests received by the original application, responses provided by the original application; and interactions between the original application and external systems.
6. The computer-implemented method of claim 1 , wherein generating the log template includes mining parameterized templates from the set of application logs to represent requests, responses, and interactions as an instance of one of the parameterized templates.
7. The computer-implemented method of claim 1 , wherein the semantic state of the original application is represented as a semantic vector.
8. The computer-implemented method of claim 1 , wherein deploying includes providing a wrapper application that is executable on the remote computer system to implement the trained classifier and the regressor.
9. The computer-implemented method of claim 1 , wherein the sequence of prior action templates includes a request template and one or more intermediate action templates that characterize a function of the original application.
10. A computer program product for emulating an application, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a hardware processor to cause the hardware processor to:
generate a log template to match patterns in a set of application logs collected from an original application;
learn semantic state representations for the original application from the log template;
train a classifier to predict a next action template based on a sequence of prior action templates;
train a regressor to generate a parameter value for a template based on a sequence of prior action templates and a particular semantic state of the original application; and
deploy the classifier, the regressor, and the log template to a remote computer system for execution to emulate the original application.
11. A system for emulating an application comprises:
a hardware processor; and
a memory that stores a computer program which, when executed by the hardware processor, causes the hardware processor to:
generate a log template to match patterns in a set of application logs collected from an original application;
learn semantic state representations for the original application from the log template;
train a classifier to predict a next action template based on a sequence of prior action templates;
train a regressor to generate a parameter value for a template based on a sequence of prior action templates and a particular semantic state of the original application; and
deploy the classifier, the regressor, and the log template to a remote computer system for execution to emulate the original application.
12. The system of claim 11 , wherein the classifier is a recurrent neural network model.
13. The system of claim 12 , wherein the recurrent neural network model is implemented as a long-short term memory (LSTM) model or a transformer model.
14. The system of claim 11 , wherein the computer program further causes the hardware processor to collect the set of application logs from the original application during operation of the original application.
15. The system of claim 14 , wherein the set of application logs include requests received by the original application, responses provided by the original application; and interactions between the original application and external systems.
16. The system of claim 11 , wherein the computer program further causes the hardware processor to mine parameterized templates from the set of application logs to represent requests, responses, and interactions as an instance of one of the parameterized templates.
17. The system of claim 11 , wherein the semantic state of the original application is represented as a semantic vector.
18. The system of claim 11 , wherein the sequence of prior action templates includes a request template and one or more intermediate action templates that characterize a function of the original application.