The fields of Artificial intelligence and Machine leaving are rapidly advancing, thanks to their incredible capabilities and use cases in almost every industry. With the increasing popularity and integration of AI into different fields, there are also problems and limitations associated with it. Root cause analysis (RCA) is a method for discovering the root causes of issues in order to find the best solutions for them. It helps in identifying the underlying reasons for incidents or failures in a model. In domains including IT operations, telecommunications, and specifically in the field of AI, the model’s increased complexity frequently results in events that reduce the dependability and effectiveness of production systems. With the help of RCA, the method looks for several factors and establishes their causal links in an effort to offer explanations for these instances.
Recently, a team of researchers from Salesforce AI has introduced PyRCA, an open-source Python Machine Learning library designed for Root Cause Analysis (RCA) in the field of Artificial Intelligence for IT Operations (AIOps). PyRCA provides a thorough framework that enables users to independently find complex causal relationships between metrics and incident root causes. The library offers both graph building and scoring operations with a unified interface that supports a variety of widely used RCA models, along with providing a streamlined method for quick model creation, testing, and deployment.
This holistic Python library for root cause analysis provides an end-to-end framework encompassing data loading, causal graph discovery, root cause localization, and RCA result visualization. It supports multiple models for creating graphs and rating root causes and helps users quickly load pertinent data and identify the causal connections between various system components. PyRCA comes with a GUI dashboard that makes interactive RCA easier, thus offering a more streamlined user experience and better aligning with real-world conditions. The GUI’s point-and-click interface has been made intuitive in nature, and the dashboard empowers users to interact with the library and inject their expert knowledge into the RCA process.
With PyRCA, engineers and researchers can now easily analyze the results, visualize the causal linkages, and move through the RCA process with the help of the GUI dashboard. Some of the key features of PyRCA shared by the team are as follows ?
- PyRCA has been developed to offer a standardized and highly adaptable framework for loading metric data with the popular pandas.DataFrame format and benchmarking a diverse set of RCA models.
- Through a single interface, PyRCA provides access to a variety of models for both discovering causal networks and locating underlying causes. Users also have the choice to completely customize each model to suit their unique requirements with models including GES, PC, random walk, and hypothesis testing.
- By incorporating user-provided domain knowledge, the RCA models offered in the library can be strengthened, making them more resilient when dealing with noisy metric data.
- By implementing a single class that is inherited from the RCA base class, developers can quickly add new RCA models to PyRCA.
- The PyRCA package provides a visualization tool that enables users to compare multiple models, review RCA results, and quickly include domain knowledge without the need for any code.
The team has explained the architecture and major functionalities of PyRCA in the technical report in detail. It provides an overview of the library’s design and its core capabilities.