"Data Mining" is one of those terms that can have either good or bad connotations. The "good" side is that lots of companies have huge databases, and data mining can be an indispensable statistical skill for culling important fcats and insights from the data -- insights that could potentially lead to a competitive advantage, or to the recognition of a critical relationship.
A recent Wall Street Journal article -- "Data Mining Isn't a Good Bet for Stock-Market Predictions" -- highlights the potential negative side of data mining. As I frequently mention in classes, "correlation" is one thing, but "causation" is something else. A relationship discovered in financial, economic, and insurance data should make sense, and should have an intuitive explanatory basis, before being used as a foundation for predictive modeling.
- Rick
No comments:
Post a Comment