USES OF FUZZY TEMPORAL DATABASES ON INFORMATION
SYSTEMS
José Enrique Pons FríasOlga Pons Capote
Department of Computer Science and Artificial Intelligence
University of Granada
PRESENTATION’S STRUCTURE
Conclusions
Uses and Commercial Systems
Background concepts
Motivation
MOTIVATION
• The storage of time-dependent objects in a database has an impact on the consistency in the database.• Humans manage time with vagueness and imprecision. There are the following types of information:
• Uncertainty: The temporal specification for an event is uncertain. Typically in temporal specification for historical databases.• Vagueness: Very often in changes of granularity.• Subjectivity or Ambiguity: The interpretation for the temporal specification relies on the hearer.
• Time Intervals as the basic unit for the representation of time in Artificial Intelligence and language understanding applications.
Three main features:
• Representation of uncertainty.• Flexible querying.• Manipulation of uncertain data.
Formal Base
Possibility theory and fuzzy sets.
Uncertainty in the price:
#ID Name Price Cuisine
1 Pizza Mario
Middle-price
Italian
2 Amadeus Expensive Brasserie
3 Wu’s house
Cheap Chinese
Motivation
Fuzzy Databases
Querying fuzzy databases:
Flexible querying
The result set is ordered with respect to a fulfillment degree.
Motivation
Fuzzy Databases
Degree #ID Name Price Cuisine
0.5 1 Pizza Mario Middle-price
Italian
0.2 3 Wu’s house Cheap Chinese
0 2 Amadeus Expensive Brasserie
The user wants a cheap restaurant with Italian cuisine
MotivationUncertainty, VaguenessVery often, the time period in which an object is valid is not known precisely.
Sometimes, the starting, the ending or both points are not known exactly.
Also, the granularity in which a time point is specified, is a source of imprecision.
People work with time in a flexible way. The user wants to specify a query to the temporal database with imprecision.
Days granularity
Months granularity
My Birthday is by May
Jan Feb Mar Apr May Jun Jul Aug Sep
Temporal specifications become vague when we
change to a finner granularity
BACKGROUND CONCEPTS
• Temporal databases• Classification of temporal databases
Background ConceptsA temporal database is a database which stores the time-dependent objects in its schema.
A relational database provides user-defined time which is an non-interpreted value.
There are three types of time despite of used-defined time:•Transaction Time, TT.•Valid Time, VT.•Decision Time, DT.
Time interval [S,E]
S E
#ID Name Age Works for
Start year
End year
1 Peter 24 John 2010 -
2 Maria 34 John 2001 -
3 John 52 - 1999 -
4 Sarah 29 Maria 2005 2009
Background ConceptsTransaction Time TTIt is the time when the fact is stored in the database.
Usually stored as a timestamp value.
#ID Name Age Works for Stored
1 Peter 24 John 15/10/2010
2 Maria 34 John 13/09/2001
3 John 52 - 10/10/1999
4 Sarah 29 Maria 05/08/2005
#ID Name Age Works for Decided
1 Peter 24 John 15/10/2010
2 Maria 34 John 13/09/2001
3 John 52 - 10/10/1999
4 Sarah 29 Maria 05/08/2005
Background ConceptsDecision Time DT
It is the time when the fact was decided to happen.
It is usually stored as a timestamp value.
Background ConceptsValid TimeIt is the time period when the fact or object is valid in the modeled reality.
It is usually stored as an interval.
Imprecision may arise in the starting, ending or both points for the valid time intervals.
Several proposals for dealing with imprecision in a valid time database.
#ID Name Age Works for
Start year
End year
1 Peter 24 John 2010 -
2 Maria 34 John 2001 -
3 John 52 - 1999 -
4 Sarah 29 Maria 2005 2009
Background ConceptsRepresentation of uncertainty in timePossibility Theory as the basis for the representation.
Several representations and proposals:
One thing in common:The fuzzy interval as the basic time unit.
Project: Narrative Sources of the Medieval Low Countries
Possibility
1
025/06/2012 5/08/2012
25/06/2012 5/08/2012
Possibility1
0
Storage of the period of time
Storage of the starting and ending points
USES OF FUZZY TEMPORAL DATABASES
• Historical databases: Usually, historians do not agree about a fact or event that happened in the past.• Inconsistent temporal knowledge: The time specification for an event may not be specified in a consistent way.• Vague or imprecise specification of time.
Uses of fuzzy temporal databasesHistorical information is typically uncertain.
Sometimes the challenge is to combine and manage all the temporal information.
The document was written duringthe pontificate of pope Alexander
Alexander II Alexander III
Possibility1
0
The document was written after king Henri IPossibility
1
0Henri I
Uses of fuzzy temporal databasesIt is possible to model inconsistent temporal information if the overlapping in the time for different versions of the same object are allowed.
#ID #Version Man. Model Start year
End year
1 001 Peugeot
207 2006 2008
2 002 Peugeot
207 2007 2011
3 001 BMW 3 2005 2008
4 002 BMW 3 2007 2012
Uses of fuzzy temporal databasesQuerying in a flexible way:
• Specifying some vagueness in the query: e.g. ‘around Christmas’.• The user is interested in a fuzzy version of the Allen’s operators: e.g. ‘before 2012’.
The user wants to obtain the cars that have beenmanufactured approximately during the followingperiod:
Possibility
1
02005 2009
Degree #ID #IID Man. Model Start year
End year
1 1 001 Peugeot 207 2006 2008
1 3 001 BMW 3 2005 2008
0.5 2 002 Peugeot 207 2007 2011
0.4 4 002 BMW 3 2007 2012
COMMERCIAL SYSTEMS
Implementations
Name Time managed Comments
Oracle Workspace Manager
Valid and Transaction time.
Package for Oracle.
TimeDB Valid and transaction time.
Interface for Oracle.
Posgree Temporal
Valid time Package for PosgreeSQL
Teradata Valid and transaction time.
Used for data-mining.
Some of the most popular implementations in the market.
CONCLUSIONS AND FURTHER WORK
Conclusions
• Temporal databases are useful in IT.
• The representation of imprecision is necessary for the user.
• Some interesting implementations, but not a truly standard.
• Further work in a flexible implementation in fuzzy temporal databases.
Thank you, questions?
José Enrique [email protected]