Database Processing for BIS Questions and Answers

Answer: B

Correct answer: Process and record transactions only

Recording transactions is the work of an operational system, not a business intelligence one. A BI system reads data that operational systems have already recorded and turns it into information for decision making. The two are deliberately separated so that heavy analytical queries do not compete with transaction processing for the same resources.

Answer: A

Correct answer: Non-duplicated data

Duplicated data is a problem; non-duplicated data is what you want. The genuine difficulties with operational data are that it is dirty, that values are missing, that it is not integrated across sources, that there is too much of it, and that it may be in the wrong format or of inconsistent granularity.

Answer: B

Correct answer: False

Sophisticated mathematical and statistical calculation is the defining characteristic of data mining, not of reporting. A reporting system uses simple arithmetic - sums, averages, counts, percentages - applied to filtered and grouped data. Keeping the definitions apart is the point of dividing business intelligence into two categories at all.

Answer: C

Correct answer: Has ordered recently, orders frequently, but doesn't place a large order when she orders

Each digit is a fifth of the customer base, with 1 the best and 5 the worst on that measure. So 1 for recency means she ordered very recently, 1 for frequency means she orders very often, and 5 for monetary value means her orders are among the smallest. A frequent, current, low-value customer - worth keeping but not worth a large sales effort.

Answer: A

Correct answer: True

RFM scores customers on recency, frequency and monetary value using nothing more than sorting, grouping and simple arithmetic, which places it in reporting rather than data mining. That is the whole basis of the classification: reporting uses simple operations on existing data, data mining uses statistical and mathematical techniques to predict.

Answer: B

Correct answer: False

A data warehouse is optimised for the opposite of transaction processing. Operational databases are tuned for many small fast writes and are normalised to protect integrity; a warehouse is tuned for large analytical reads, is usually denormalised into star schemas, and is loaded in batches rather than updated continuously.

Answer: B

Correct answer: Process transactions

BI systems obtain data by reading an operational database directly, by processing extracts taken from one, or by processing data bought from a vendor. What they do not do is process transactions themselves - that is the operational system’s job, and keeping the two apart is what protects transaction performance from analytical load.

Answer: C

Correct answer: A subset of a data warehouse serving a single department or subject area

A data mart narrows the warehouse to the data one group actually needs, which makes queries faster and access control simpler. It may be carved out of an existing warehouse or built first and later consolidated into one.

Answer: A

Correct answer: both standard SQL and extensions to SQL

Much reporting is straightforward SQL - selecting, filtering, grouping and aggregating - and RFM analysis, for instance, can be done entirely that way. But OLAP reporting needs the analytical extensions that support cubes, rollups and drill-down, which standard SQL does not provide. Reporting therefore uses both, which is why products offer the extensions at all.

Answer: D

Correct answer: confidence

Confidence is the conditional probability of one item given another - the probability that Video D is rented given that Video C has been. Support is the unconditional probability of the pair occurring together, and lift is the ratio of confidence to the base rate of D, which shows whether C actually makes D more likely.