34
Contents Motivation Context Proposal Comparison Conclusions and . . . Home Page Title Page JJ II J I Page 1 of 34 Go Back Full Screen Close Quit A Possibilistic Valid-time Model Jos´ e Enrique Pons 1 Christophe Billiet 2 Olga Pons Capote 1 Guy De Tr´ e 2 1 Department of Computer Science and Artificial Intelligence University of Granada, Spain {jpons,opc}@decsai.ugr.es 2 Department of Telecommunications and Information Processing Ghent University, Belgium {Christophe.Billiet,Guy.De.Tre}@telin.ugent.be June 29, 2012

A possibilistic Valid time model

Embed Size (px)

DESCRIPTION

Information in databases can be imperfect and this imperfection has several forms and causes. In some cases, a single value should be stored, but it is (partially) unknown. The uncertainty about which value to store leads to the aforementioned imperfection. In temporal databases, uncertainty can arise, concerning which temporal notion needs to be stored. Because in temporal databases, temporal notions influence the consistency with which the database models the reality, this uncertainty has a direct impact on the consistency of the model. To represent this temporal uncertainty, previous works have adapted fuzzy sets with conjunctive interpretation, an approach that might prove misleading. This work presents a model that represents the uncertainty using possibility and necessity measures, which are fuzzy sets with disjunctive interpretations.

Citation preview

Page 1: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 1 of 34

Go Back

Full Screen

Close

Quit

A Possibilistic Valid-timeModel

Jose Enrique Pons1 Christophe Billiet2 Olga Pons Capote1

Guy De Tre2

1 Department of Computer Science and Artificial IntelligenceUniversity of Granada, Spain{jpons,opc}@decsai.ugr.es

2 Department of Telecommunications and Information ProcessingGhent University, Belgium

{Christophe.Billiet,Guy.De.Tre}@telin.ugent.be

June 29, 2012

Page 2: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 2 of 34

Go Back

Full Screen

Close

Quit

1. Contents

The structure of the presentation is:

2 Motivation.

3 Context:

3.1 Temporal databases.

3.2 Possibilistic variables and fuzzy numbers.

4 Proposal: Interval evaluation by ill-known con-straints.

5 Analysis of proposed transformations.

6 Conclusions and future work.

Page 3: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 3 of 34

Go Back

Full Screen

Close

Quit

2. Motivation

• The study of fuzzy intervals is of particular in-terest in temporal database research.

• To optimize the storage of fuzzy temporal inter-vals, some transformations have been proposed.⇒ Information Lost.

• The proposal is a framework to deal with theevaluation of ill-known temporal intervals.

Page 4: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 4 of 34

Go Back

Full Screen

Close

Quit

I Before J I

J

I Equal J

-Time

J

I Meets J J

I Overlaps J J

I During J J

I Starts J

I Finishes J

J

J

Relations

Page 5: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 5 of 34

Go Back

Full Screen

Close

Quit

3. Context

3.1 Temporal databases

3.2 Possibilistic variables and fuzzy numbers

Page 6: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 6 of 34

Go Back

Full Screen

Close

Quit

3.1. Temporal Databases:

A temporal database is a database that managesthe time in its schema.

• The time is usually represented as an interval inthe database.

X Y

• The user provides a crisp temporal interval in thequery specification.

Page 7: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 7 of 34

Go Back

Full Screen

Close

Quit

Example:

ID Name Year Works for Start Year End Year1 Peter 1987 John 2010 -2 Maria 1978 John 2001 -3 John 1954 - 1999 -4 Sarah 1982 Maria 2005 2009

Consider that ID is the primary key.Problem: If Sarah is hired in 2010, we can notinsert the new tuple.

Page 8: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 8 of 34

Go Back

Full Screen

Close

Quit

Example:

New primary key:

{ ID ∪[ Start Year, End Year ]}

ID Name Year Works for Start Year End Year1 Peter 1987 John 2010 -2 Maria 1978 John 2001 -3 John 1954 - 1999 -4 Sarah 1982 Maria 2005 20094 Sarah 29 Maria 2010 -

Also a consistence mechanism must be defined...

Page 9: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 9 of 34

Go Back

Full Screen

Close

Quit

Example:

Some spurious values might be inserted:

ID Name Year Works for Start Year End Year1 Peter 1987 John 2010 -2 Maria 1978 John 2001 -3 John 1954 - 1999 -4 Sarah 1982 Maria 2005 20094 Sarah 29 Maria 2001 20074 Sarah 1982 Maria 2010 -

Usually, DML sentences (insert, update, delete) arere-defined to ensure consistency.

Page 10: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 10 of 34

Go Back

Full Screen

Close

Quit

3.1.1. Valid-time DML

ID Entity Start End3 E.U. 15/3/2012 -

• Modify.

• Insert.

• Delete.

Page 11: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 11 of 34

Go Back

Full Screen

Close

Quit

Modify:

ID Entity Start End3 E.U. 15/3/2012 30/3/20123 E.U. 4/4/2012 UC

• Insert new information about an existing entity.

• This operation does not remove any previous

value for the entity.

Page 12: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 12 of 34

Go Back

Full Screen

Close

Quit

Insert:

ID Entity Start End3 E.U. 15/3/2012 30/3/20124 N.A.T.O. 25/3/2012 4/4/20123 E.U. 4/4/2012 UC

There are two main cases:

1. The entity is not in the relation. E.g., entity with

ID = 4.

2. The entity is already in the relation. E.g. entity

with ID = 3.

Page 13: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 13 of 34

Go Back

Full Screen

Close

Quit

Insert:

ID Entity Start End3 E.U. 15/3/2012 30/3/20124 N.A.T.O. 25/3/2012 4/4/20123 E.U. 4/4/2012 11/6/20123 E.U. 12/6/2012 -

If the entity is already in the relation, then:

• Insert the new version for the entity if it does not

overlap any other version.

• Modify and close the current version of the entity

and insert the new version.

• Reject the insertion if the time interval for the

entity does overlap any existing valid-time for the

entity.

Page 14: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 14 of 34

Go Back

Full Screen

Close

Quit

Delete:

ID Entity Start End4 N.A.T.O. 25/3/2012 4/4/2012

Removes all the versions for a given entity. For ex-

ample consider the deletion of the entity with ID =

3.

Page 15: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 15 of 34

Go Back

Full Screen

Close

Quit

3.1.2. Temporal models

• Valid-Time: The time when the fact is true inthe modelled reality.

• Transaction time: The time when the fact isstored in the database.

• Decision-time: The time when an event wasdecided to happen.

• Bi-temporal: Valid and transaction time.

• Tri-temporal: Valid, transaction and decisiontime.

Page 16: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 16 of 34

Go Back

Full Screen

Close

Quit

3.1.3. Temporal operators:

Allen’s temporal relation between intervals:Operator Inverse RepresentationX Before Y After XXX YYY

X Equal Y Equal XXXY Y Y

X Meets Y Meet by XXXYYY

X Overlaps Y Overlap by XXXY Y Y

X During Y Contains XXXY Y Y Y Y Y

X Starts Y Start by XXXY Y Y Y Y Y

X Finishes Y Finished by XXXY Y Y Y Y Y

Page 17: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 17 of 34

Go Back

Full Screen

Close

Quit

3.2. Possibilistic variables and fuzzy

numbers

Two different natures for a fuzzy set:

• Conjunctive nature: The fuzzyfication of aregular set. This interpretation corresponds withthe following two semantics: Degree of prefer-ence and degree of similarity.

• Disjunctive nature: In this case, the disjunc-tive nature indicates a description of incompleteknowledge. This interpretation corresponds withthe semantics for the degree of uncertainty.

Page 18: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 18 of 34

Go Back

Full Screen

Close

Quit

Possibilistic Variable:A possibilistic variable X over a universe U is de-fined as a variable taking exactly one value in U ,but for which this value is (partially) unknown.The possibility distribution πX gives the availableknowledge about the value that X takes. For eachu ∈ U , πX(u) represents the possibility that Xtakes the value u.

-

6

1

N1 2 3 4

r

r

πX

Page 19: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 19 of 34

Go Back

Full Screen

Close

Quit

It is important to understand the difference betweenthe following two concepts:

• A possibilistic variable X is bounded to takeonly one value , but this value is not known dueto incomplete knowledge.

• An ill-known set : a possibilistic variable definedover the universe P(U).

Note that while a possibilistic variable refers to one(partially) unknown value, an ill-known set is a crispset but, for some reason, (partially) unknown.

Page 20: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 20 of 34

Go Back

Full Screen

Close

Quit

Fuzzy numbers and fuzzy intervalsA fuzzy interval is a fuzzy set M on the set of realnumbers R such that:

∀(u, v) ∈ R2 :

∀w ∈ [u, v] : µM(w) ≥ min(µM(u), µM(v))

∃m ∈ R : µM(m) = 1

If m is unique, then M is referred to as a fuzzynumber, instead of a fuzzy interval.

1

0

possibility

values

D-a D D+b

1

0

possibility

α β γ δ

Page 21: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 21 of 34

Go Back

Full Screen

Close

Quit

4. Interval evaluation by ill-known constraints

4.1 Constraint

4.2 Ill-known constraint

4.3 Example

Page 22: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 22 of 34

Go Back

Full Screen

Close

Quit

4.1. Constraint:

Given a universe U , a constraint C on a set A ⊆ Uis specified by means of the binary relation R ⊆ R2

and a fixed value x ∈ U :

C4= (R, x)

It is said that a set A satisfies the constraint C ifand only if:

∀a ∈ A : (a, x) ∈ R.

Page 23: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 23 of 34

Go Back

Full Screen

Close

Quit

4.2.

Ill-known constraint:

Given a universe U , an ill-known constraint C on a set A ⊆ U is specified bymeans of a binary relation R ⊆ U2 and an ill-known value X, i.e.:

C4= (R,X) .

The uncertainty that a set A ⊆ U satisfies C is given by:

Pos(C(A)) = mina∈A

(Pos(a,X) ∈ R

)= min

a∈A

(sup

(a,w)∈R

πX(w)

)Nec(C(A)) = min

a∈A

(Nec(a,X) ∈ R

)= min

a∈A

(inf

(a,w)/∈R1− πX(w)

)

Page 24: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 24 of 34

Go Back

Full Screen

Close

Quit

4.3.

Consider the two ill-known values X and Y .

X Y

Page 25: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 25 of 34

Go Back

Full Screen

Close

Quit

Allen Relation Constraints B(C1(I), ..., Cn(I)

)I before J C1

4= (<,X) C1(I)

I equal J

C14= (≥, X) C1(I) ∧ ¬C2(I) ∧ C3(I) ∧ ¬C4(I)

C24= (6=, X)

C34= (≤, Y )

C44= (6=, Y )

I meets JC1

4= (≤, X) C1(I) ∧ ¬C2(I)

C24= (6=, X)

I overlaps JC1

4= (<, Y ) C1(I) ∧ ¬C2(I) ∧ ¬C3(I)

C24= (≤, X)

C34= (≥, X)

I during J

C14= (>,X)

(C1(I) ∧ C2(I)

)∨(C3(I) ∧ C4(I)

)C2

4= (≤, Y )

C34= (≥, X)

C44= (<, Y )

I starts JC1

4= (≥, X) C1(I) ∧ ¬C2(I)

C24= (6=, X)

I finishes JC1

4= (≤, Y ) C1(I) ∧ ¬C2(I)

C24= (6=, Y )

Page 26: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 26 of 34

Go Back

Full Screen

Close

Quit

5. Analysis of proposedtransformations

Optimize storage⇒ Transformation from two fuzzynumbers to a fuzzy interval.

2 main proposals:

• Transf. Preserving the imprecision.

• Transf. based on the convex-hull.

Drawbacks:

• (Dubois and Prade): the fuzzy interval is apossibility distribution on R while the twofuzzy numbers are a set that belong to P(R).

• The lack of the necessity measure, used forranking purposes.

Page 27: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 27 of 34

Go Back

Full Screen

Close

Quit

5.1. Transformation that preserves the

imprecision

1

0

possibility

ds-as ds

1

0

possibilityds+bs

de-ae de de+be

S1

S2

S3

S4

Page 28: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 28 of 34

Go Back

Full Screen

Close

Quit

5.2. Transformation based on the convex

hull

1

0

possibility

1

0

possibility

ds deds-as ds+bs de-ae de+be

ds-as ds de de+be

Page 29: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 29 of 34

Go Back

Full Screen

Close

Quit

ComparativeConsider two ill-known points representing a timeinterval: X = [3, 2, 1] and Y = [7, 2, 3]The value for I = [a, b] is [3, 6]The relation R: I is inside X :

Method Possibility NecessityIll-known constraint 1 0.5

Preserving the imprecision 0.667 -Convex hull 1 -

Nec (C (A)) > 0⇐⇒ Pos (C (A)) = 1

Page 30: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 30 of 34

Go Back

Full Screen

Close

Quit

Page 31: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 31 of 34

Go Back

Full Screen

Close

Quit

Pos+Nec

0

1

2

Poss Nec

Page 32: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 32 of 34

Go Back

Full Screen

Close

Quit

6. Conclusions

• The necessity measure is lost when dealing witha transformation.

• The possibility measure in the transformations is(w.r.t. the ill-known evaluation):

– Convex hull returns the same value as possi-bility.

– The preserving the imprecision approach re-turns a different value.

• If the support for the ill-known values do overlap,it is not possible to compute any transformations.

Page 33: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 33 of 34

Go Back

Full Screen

Close

Quit

Future work:

• A new theoretical model for valid-time databases.

• Extension of the Allen’s relations for the compar-ison between two ill-known values.

• Implementation of the theoretical model in a re-lational database.

Page 34: A possibilistic Valid time model

Contents

Motivation

Context

Proposal

Comparison

Conclusions and . . .

Home Page

Title Page

JJ II

J I

Page 34 of 34

Go Back

Full Screen

Close

Quit

Thank you!

Questions?

Contact:

[email protected]

http://decsai.ugr.es/˜ jpons