Next: Commercial DBMS and
Up: Data Mining
Previous: Data Classification
Time series applications have data as a sequence of events, with each event
having a time of occurance. These type of applications finds prominent use in
financial analysis and analysis by retailers. Financial analysts frequently try
to
predict interest rate fluctuations or stock performance based on a series of
preceding events. Time series analysis use temporal aggregation
[ZM97], details of which are given in next paragraph.
A few kind of temporal aggregation used for time-series analysis are:
- Running aggregates : Aggregate computed for each point in the sequence,
based on all the earlier events, e.g. price of a mutual fund from its
inception.
- Moving-window aggregates: Aggregate computed based on the last n events
in the sequence, e.g. average price of a stock since quarterly dividend.
- Temporal grouping: The time line is partitioned into consecutive time
intervals and aggregate computed for each interval, based on events in the
interval, e.g. price of a stock average over month, quarter, year.
As we will see user-defined aggregates can be used to express temporal
aggregates, under the assumption that the data is sorted according to time.
Punit Bhargava
Wed Mar 11 18:50:53 PST 1998