Topic 8 · Databases (HL Option)

Relational (tables + keys) · Object-oriented · Network · Spatial · Multi-dimensional (cubes).

Online Transaction Processing. Operational DB. Many small INSERT/UPDATE. Live data. Normalised (3NF).

Online Analytical Processing. Data warehouse. Aggregate queries on historical data. Often denormalised.

Subject-oriented · Integrated (many sources) · Time-variant (historical) · Non-volatile (read-mostly).

Extract (pull from sources) → Transform (clean, standardise) → Load (insert into warehouse). Usually nightly.

Modern variant: Extract → Load (raw) → Transform inside warehouse. Cheaper, more flexible.

Group similar items WITHOUT labels (unsupervised). E.g. customer segments.

Predict KNOWN labels (supervised). E.g. spam / not-spam, fraud / legit.

Find rules 'A → B'. E.g. 'bought bread → bought butter'. Market-basket analysis.

Predict NUMERIC future from historical patterns. E.g. sales next month.

Find outliers / anomalies. E.g. unusual credit-card spending.

Business Intelligence tools query the warehouse for reports & dashboards.