kdd

Knowledge Discovery in Databases (KDD) is the nontrivial process of identifying valid, novel, potentially useful, and ultimately understandable patterns in data. It refers to the overall cycle of steps used to turn raw data into meaningful knowledge, typically involving data selection, cleaning, integration, transformation, data mining, evaluation, and presentation. In this view, data mining is the core analytical step that employs algorithms to extract patterns, while the surrounding steps ensure data quality and the interpretability of results.

The concept emerged from research in databases and machine learning during the 1980s and 1990s. A widely cited formulation was published in 1996 by Fayyad, Piatetsky-Shapiro, and Smyth, who described KDD as an iterative process that culminates in the interpretation and evaluation of discovered patterns. Since then, KDD has become a foundational term in data science, data mining, and artificial intelligence, and it informs practical methodologies for handling large and complex data sets.

Typical activities within KDD include selecting relevant data sources, preprocessing to address missing values and inconsistencies, transforming data into suitable representations, applying data mining algorithms (such as classification, clustering, association, or anomaly detection), and validating the results against domain knowledge. Deployment involves integrating findings into decision-making, systems, or business processes.

The field is associated with a professional community and venues such as the ACM SIGKDD Conference on Knowledge Discovery and Data Mining, which showcases research advances and applications. Related concepts include data mining, machine learning, and data science, as well as practical frameworks like CRISP-DM.