Latha M N
#ATAITF: A Festival of AI Testing Tools & Techniques
About Speaker
Latha M N
Programmer Analyst
Cognizant Technology Solutions
Programmer Analyst at Cognizant Technology with 2.4 years of experience in performance testing. Skilled in analyzing and optimizing application performance using tools like JMeter and LoadRunner, with a strong interest in leveraging AI-driven solutions to enhance efficiency and accuracy in performance testing.
Anomaly Detection in Performance Metrics using AI explores the integration of machine learning techniques to identify performance bottlenecks, predict system failures, and enhance test efficiency. Traditional performance testing methods rely on manual analysis and predefined thresholds, which may overlook subtle anomalies. This approach applies AI to detect deviations in performance data, providing real-time insights and proactive alerts. By incorporating AI-driven anomaly detection, performance testing can become more efficient, scalable, and adaptive, leading to faster issue resolution and improved system reliability.
More Speakers
Topic – Anomaly Detection in Performance Metrics Using AI
Performance testing is crucial for ensuring system reliability, scalability, and efficiency. However, the manual analysis of performance metrics during testing can be time-consuming and prone to human error. This talk introduces an AI-driven approach for anomaly detection in performance testing metrics, enabling testers to identify performance bottlenecks and unusual patterns with higher precision and speed.
The session will focus on using machine learning models to monitor key performance metrics such as response time, CPU/memory utilization, throughput, and error rates. Attendees will learn how unsupervised algorithms (e.g., Isolation Forest, DBSCAN) and time-series models (e.g., ARIMA, LSTM) can detect outliers or anomalies in real-time.
The presentation will demonstrate:
Collecting and preprocessing performance data from tools like JMeter, Dynatrace, or Prometheus.
Training AI models on historical performance data to establish baselines.
Real-time anomaly detection with AI tools integrated into CI/CD pipelines.
Automatically generating alerts for identified anomalies and correlating them to potential root causes.
This AI-based approach reduces manual effort, enhances scalability, and ensures faster identification of performance issues, significantly improving testing efficiency.
Attendees will walk away with actionable insights and practical steps to implement AI for anomaly detection in their performance testing workflows.