A grocery chain might mine millions of receipts and find which products are often bought together, or that one store’s sales drop every time a nearby road closes. Nobody asked those exact questions in advance. The output is a pattern, a segment or a hypothesis that merchandising, pricing or operations teams then test.
The 1996 overview by Usama Fayyad and colleagues frames mining as one step in “knowledge discovery in databases”: choose the data, clean it, transform it, mine it and then interpret what comes out. Most of the effort sits around the mining step. Machine learning often does the mining, through methods such as cluster analysis, but so can SQL summaries and charts. Building and serving one prediction model is a narrower job.
Search enough data and some patterns will appear by chance. Treat a mined pattern as a lead to confirm on fresh data, not as proof. Data mining also isn’t data federation: federation is about querying data where it lives, mining is what you do once you can reach it.
Sources
- Fayyad, Piatetsky-Shapiro and Smyth: From Data Mining to Knowledge Discovery in Databases (AI Magazine, 1996) — Overview article that places data mining as one step within the broader knowledge discovery in databases (KDD) process of selecting, preparing, mining and interpreting data; checked October 24, 2026.