Automated testing of software applications using machine learning edited
0 likes191 views
Machine Learning is the next internet. It is the backbone of search engines, driverless car, paperless banking, and facial recognition in forensics. Running automated software tests with lesser human intervention without the risk of schedule delays is now a reality. This presentation will explore several practical machine learning concepts that are being adopted to test software applications.
![©Unpickle, 2016. All Rights Reserved - Privileged and Confidential
Adaptive Automation – Test Case Generation
15
• Adaptive Automation: Leveraging Machine Learning to Support Uninterrupted Automated Testing of Software
Application; Rajesh Mathur, Scott Miles, Miao Du arXiv:1508.00671v1 [cs.SE] 4 Aug 2015
1. Automatically exploring the screens or
pages
2. Generating tests which exercise all of the
discovered fields
3. Guide the values used in the test cases
4. Employ a database of common field
names and associated formats
5. Historical user interaction logs to test
most often used navigation paths
Check that it
only accepts
mobile formats](https://image.slidesharecdn.com/automatedtestingofsoftwareapplicationsusingmachinelearning-edited-160822082948/85/Automated-testing-of-software-applications-using-machine-learning-edited-15-320.jpg)
More Related Content
Similar to Automated testing of software applications using machine learning edited (20)
Recently uploaded (20)
Automated testing of software applications using machine learning edited
- 1. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidentialwww.unpickle.in Automated testing of software applications using Machine Learning Milind Kelkar Chief Analytics Officer August 05, 2016 1
- 2. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential CONFERENCE PROGRAM AUG 05, 2016 2 • STePIN SUMMIT 2016 • Conference Day 2 • Park Plaza, Bangalore, India
- 3. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Today’s Agenda 3 • What is Machine Learning (ML) • Framework to leverage ML • Software Testing Use Cases
- 4. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Support Vector Machine Machine Learning Techniques 4 Deep Learning Dimension Reduction Regression Anomaly Detection Classification Ensemble Bayes
- 5. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Deep Learning 5 ww.huffingtonpost.com Driverless Car using Neural Networks Stay on Course Accelerator Brakes Steering Wheel
- 6. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential 6 Zebra or Horse Stripes Yes SLP MLP ABT No GBN PNN Classification
- 7. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Why Machine Learning for Testing 7 • Embed a “robot” inside a program – Fast, consistent conformance of a product to requirements • Real time decisioning – Reduce time to test and monitor • Software applications we cannot program by hand – Analyse the weaknesses of test suites so as to be able to iteratively improve them • Self-Customization Programs – Auto-correct complexity of the software application
- 8. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential T ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Framework to leverage ML 8
- 9. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Leveraging Machine Learning (ML) 9 Machines testing software applications Machines masquerading as Humans
- 10. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Summarizing Machine Learning 10 • Virtual Assistant • Re-engineering of test suites • Driverless Car • Adaptive Automation • Spam detection • Software Defect Prediction • Intelligent Search • Live Validation Humans RationalActThink
- 11. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Software defect prediction in 3 steps 11 • Goal of Machine Learning 1. BUILD computer systems 2. Learn from EXPERIENCE 3. Improve PERFORMANCE over experience 1. BUILD % of code correctly classified as “defect” or “not a defect” Machine classifying a line code as “defect” or “not a defect” 2. EXPERIENCEMachine - Learning from YOU labelling a code as “defect” or “not a defect” 3. PERFORM
- 12. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Software defect prediction 12 Occurrence of corrections to a tested file Physical Line of Code Total Line count Blank Line of Code Comment Line of Code Coupling between Objects Cyclomatic Complexity Number Weighted Methods per Class Depth of Inheritance Tree Source Code Metric Set Test Source Code Metric Set
- 13. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Intelligent Search 13
- 14. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Live Validation - Infer 14 • Finding Bugs before it goes Live • Potential issues and bad coding habits • Guesses rationally • Spot bugs in minutes • Fix rate of about 80 percent
- 15. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Adaptive Automation – Test Case Generation 15 • Adaptive Automation: Leveraging Machine Learning to Support Uninterrupted Automated Testing of Software Application; Rajesh Mathur, Scott Miles, Miao Du arXiv:1508.00671v1 [cs.SE] 4 Aug 2015 1. Automatically exploring the screens or pages 2. Generating tests which exercise all of the discovered fields 3. Guide the values used in the test cases 4. Employ a database of common field names and associated formats 5. Historical user interaction logs to test most often used navigation paths Check that it only accepts mobile formats
- 16. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Act like Humans 16 Commercial Bank of Dubai https://www.cbd.ae/ Virtual Assistant on the Website
- 17. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Link Problems to Test Cases 17 • Using Machine Learning to Refine Black-Box Test Specifications and Test Suites; Lionel C. Briand ; Yvan Labiche; Zaheer Bawar; (2007) Problems Causes Misspecification Missed Category Unused Categories Ill-defined Choices Missing Test Cases
- 18. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Languages for Machine Learning 18 • Prolog • Python • R • Matlab • Scala • Clojure • Ruby • Rapidminer • Java • Weka Toolkit • SpringXD • Mahout
- 19. ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential T ©Unpickle, 2016. All Rights Reserved - Privileged and Confidential Thank You 19