[ Machine Learning & Artificial Intelligence - Machine Certified Test Data - Reality (or) Hype?? ]
With a universal shift towards extreme automation, every IT specialist has been preached on adopting “Duality of Automation and Innovation” on every aspects of the day to day tasks (or) activities that are typically carried out by them. Test Data Management is in no way insulated from this cultural change. This has set the stage for the industry to look at leveraging emerging areas such as “Machine Learning”, “Artificial Intelligence” & “Deep Learning” to bring in some innovation and efficiency to their projects. This paper would throw light on how Machine Learning Algorithms could help us in evaluating Manufactured Test Data to ensure completeness and correctness. Once we tame and train the algorithms on the expected data patterns, the test data sets can be subjected to analysis and any deviations with the expected patterns are raised as anomalies, by the algorithms. There is no need for testers to understand the rules/requirements on test data and code them. This is the USP of leveraging Machine Learning Algorithms.
There are 2 major challenges here:
- Identifying the expected data set pattern for training the algorithm
- Tuning the algorithm to 100% accuracy on learning the expected data set
Real time scenarios have been handpicked for illustrating how the above 2 challenges can be addressed. Once we address this challenges, rest becomes child play. Machine Learning Algorithms are rendered through languages like Java, Python & R. Illustrations in this paper would be through Java.
By extrapolating the ides to respective real time scenarios, the idea of leveraging Machine Learning for Test Data Management would become a reality. Intended audience for this paper would be Test Engineers, Test Leads, Test Managers, Test Architects & Test Consultants.
By extrapolating the above ides to respective real time scenarios, the idea of leveraging Machine Learning for Test Data Management would become a reality. Intended audience (who would get benefitted) for this paper would be Test Engineers, Test Leads, Test Managers, Test Architects & Test Consultants..
Muthu Venkatesh is a Principal Test Consultant , has 17+ years of experience in testing industry and has successfully strategized end to end testing for very large implementations n DWT/BIG Data for financial clients like Bank of America, HSBC, Barclays, DB, DNB, NAB, RBS,CGC,Westpac, Wells Fargo and Citibank. He also plays an active role in building and strengthening the DWT and BIG Data COEs in Infosys. He is well known for his technical and domain expertise and has been the “TO GO” person for all BIG Data Testing proposals, trainings and people enablement. He is part of the Machine Learning & Artificial Intelligence COE and is piloting the implementation of the same for data services testing.
He has developed macros and utilities for automating the repetitive validations across accounts thereby imparting cross-pollination across projects. He has won various accolades for presenting white papers in Infosys Internal forums as well as in international testing conferences like UNICOM, STC etc..
Karthikeyan Mani is a Test Lead at Infosys Ltd. He has a Bachelor of Engineering Degree in Computer Science. He is having 10+ years of Software Testing experience in Investment Banking (Credit Risk, Collateral & Wealth Mgmt, and Portfolio Reconciliation & Dispute Mgmt), Telecom (Order Mgmt) and Logistic (Airline & Shipping Cargo) domains. He has been active in Data Analysis, Data Migration, Data warehouse and Business Intelligence, Environment and Test Management, SIT, UAT and Database testing projects. His international experience includes Data Analysis in an integrated, cross-asset, front–to-back position Mgmt, pricing, risk Mgmt platform for a leading US banking client.