Previous

🏠 Home

Next

Temporal database

Description

A temporal database is a collection of transactions ordered by their timestamp. A sample temporal database generated from the set of items, I={a,b,c,d,e,f}, is shown in below table:

TID	Timestamp	Transactions
1	1	a, b, c
2	2	d, e
3	4	a, e, f
4	7	d, f, g

Rules to create a temporal database

1. Since TID of a transaction implicitly represents the row number, this information can be ignored to save space.
2. The first column in the database must represent a timestamp.
3. The timestamp of the first transaction must always start from 1. The timestamps of remaining transactions follow thereafter. In other words, the timestamps in a temporal database must be relative to each other, rather than being absolute timestamps.
4. Irregular time gaps can exist between the transactions.
5. Multiple transactions can have a same timestamp. In other words, multiple transactions can occur at a particular timestamp. (Please note that some pattern mining algorithms, especially variants of ECLAT, may not work properly if multiple transactions share a common timestamp.)
6. All items in a transaction must be seperated with a separator.
7. The items in a temporal database can be integers or strings.
8. ‘Tab space’ is the default seperator. However, temporal databases can be constructed using other seperators, such as comma and space.

Format of a temporal database

timestamp<sep>item1<sep>item2<sep>...<sep>itemN

An example

1   a   b   c
2   d   e
4   a   e   f
7   d   f   g