Atomic Talk Title: RealTime Defects Categorization using ML/AI

AI is significantly transforming the field of software testing, offering new tools and methodologies to enhance the efficiency, accuracy, and scope of testing activities. Triaging defects, determining their root causes, prioritizing them, and assigning them to the appropriate technical team is a time-consuming process that can be automated for real time action using AI. Urgent production incidents occur, often critical and requiring swift intervention. QA must first carefully analyze the issue, putting in a significant amount of manual effort to pinpoint the root cause, assign the correct priority, and determine if it’s a defect, an enhancement, or a task. An organization providing services to it several clients which in turn have several customers and if these customers have issues, incident emails are flooded to organization’s support team. They are required to do thorough analysis in a short span of defined SLA to respond back to customers which is complicated, time consuming and if this analysis goes wrong then it leads to email threads.

Below are the listed activities where ML can be utilized to reduce this time frame –

1. Image Classification ML model can categorize which client’s customer are raising this request , it helps us in directing ticket to internal respective development team who had implemented this service. My organization had 33 clients for which I had gathered 30 logo images for each client to train my model. After getting an accuracy of ~85% , deployed it for real time use cases.

2. Text Classification model – NLP helps in analyzing whether its an incident, defect or enhancement request. It also helps in furnishing what could be root cause.I have trained existing support tickets text, utilized web scrapping using Beautiful soup library , sentiment analysis for identifying right category of customer email input.

Also I’ll present couple of below innovative approaches for optimizing ML model performance –

1. Caching Mechanism – New image prediction takes time so we should avoid direct feeding to ML model.

2. Handle different Languages tickets – Organization customers can be in several countries and if tickets are originating from non-English speaking countries like Spanish how this utility can handle text and automatically triage it.

Talk Takeaways

• AI Impact in SoftwareTesting
• Text Scraping
• Automation Utility Utilizing Deep
• Learning ML Models