Multivariate Data Mining Oracle Tips by Burleson Consulting

High Performance Data Warehousing

Problem Solving For The Oracle Warehouse  

Multivariate Analysis

This type of analysis includes comparing various classifications of database objects. Multivariate analysis involves the use of chi-square statistical techniques to compare ranges of values among different classifications of objects. For example, a supermarket may keep a data warehouse of each customer transaction. Multivariate techniques are used to search for correlations among items, so the supermarket management can place these correlated items nearby on the shelves. One supermarket found that whenever males purchased diapers, they were also likely to buy a six-pack of beer! As a consequence, the supermarket was able increase sales of beer by placing a beer display between the diaper displays and the checkout lines.

Multivariate analysis is normally used when the answers to the query are unknown and is commonly associated with data mining and neural networks. For example, whereas a statistical analysis may query to see what the correlation is between customer age and probability of diaper purchases when the end user suspects a correlation, multivariate analysis is used when the end user does not know what correlation may exist in the data. A perfect example of multivariate analysis can be seen in the analysis of the Minnesota Multi-phasic Personality Inventory (MMPI) database. MMPI is one of the most popular psychological tests in America, and millions of Americans have taken this exam. By comparing psychological profiles of subjects with diagnosed disorders to their responses to the exam questions, Psychologists have been able to generate unobtrusive questions which are very highly correlated with a specific mental illness. One example question relates to a subject’s preference to take showers versus baths. Answers to this question are very highly correlated with the MMPI’s measure for self-esteem. (It turns out that the correlation showed that shower-takers tend to have statistically higher self-esteems than bath-takers.)

Note that the users of this warehouse do not seek answers about why the two factors are correlated; they simply look for statistically valid correlations. This approach has made the MMPI one of the most intriguing psychological tests in use today; by answering the seemingly innocuous 500 True/False questions, psychologists can gain an incredible insight into the personality of a respondent.

---

This is an excerpt from "High Performance Data Warehousing", copyright 1997.

If you like Oracle tuning, you may enjoy the book Oracle Tuning: The Definitive Reference , with over 900 pages of BC's favorite tuning tips & scripts.  You can buy it directly from the publisher and save 30%, and get instant access to the code depot of Oracle tuning scripts.