Training AI Models: Preparing and Managing AI Algorithms for AIOps
Satyanarayana Murthy Polisetty · International Journal of Scientific Research in Computer Science Engineering and Information Technology · 2023
Artificial Intelligence plays a critical role in AIOps by enhancing decision-making and reducing human intervention in IT operations. This paper dives deep into the preparation and management of AI models within the IBM Cloud Pak for AIOps framework. It focuses on how AI algorithms are trained and deployed to address specific challenges like incident detection, anomaly prediction, and service availability optimization. The paper highlights key methodologies for selecting the right AI models, preparing data, and maintaining the algorithms over time. A major section of the article is dedicated to understanding the types of algorithms available, such as natural language log anomaly detection and metric anomaly detection, and how they are fine-tuned for real-time data analysis. Novel methodologies are proposed for managing AI models, including dynamic algorithm adjustments based on operational needs, and scaling models in large, distributed environments. The article suggests a new approach where models continuously learn from incoming data, ensuring the AI remains relevant and adaptive to changing IT environments. By integrating real-time monitoring tools into the model management pipeline, it proposes that IBM Cloud Pak can dynamically allocate resources and adjust algorithm complexity as required, improving operational efficiency.