So why “Data Mining” ? (vs. Pattern Recognition or Statistics?)

• Scalability to large data sets:
• what if data is disk-resident and can only afford to scan it a few times?
• what if “random” sampling is not sufficient or even more efficient than scan? distributed system?
• EDA with large data sets and high dimensionality
• New algorithms are beginning to appear (some without proper statistical foundation)
• What is possible/impossible with massive data sets?
• Automation: effective use by non-statisticians

---