Foundations of Preferences in Database Systems

1
Foundations of Preferencesin Database Systems

• Werner Kießling

2
Motivation

• Provide a natural way to express wishes
• Preferences are personal wishes
• I like A better than B
• Cope with empty-result, flooding problems
• SELECT FROM car
• WHERE make 'Opel'
• PREFERRING (category 'roadster' ELSE
• category ltgt 'passenger'
• AND price AROUND 40000
• AND HIGHEST(power))
• CASCADE color 'red'
• CASCADE LOWEST(mileage)

3
Preferences Model

• Preference Pstrict partial orderP (A, ltP) on
  dom(A)x ltP y -gt I like y better than x
• Better-than-graph

Level 1 val1 val3
Level 2 val2 val4
Level 3 val5 val6 val7
4
Preference Engineering

• Base preferencesPOS, NEG, POS/NEG
• Numeric base preferenceAROUND, BETWEEN,
  LOWEST,SCORE
• COMPLEX preferencesPareto P1?P2, Prioritized
  P1P2, rank,
• Accumulation and Aggregation

dom(A)
POS-set
NEG-set
z
v
5
BMO query model

• Best match only
• P (A, ltP) where A A1, A2 ., Ak
• Pr (RA, ltP) database preference
• Tuple t is a perfect match, iff tA ? max(P)
  and tA ? R

6
Example

• P1 LOWEST(price)
• P2 LOWEST(mileage)
• P (price,mileage,ltP1?P2)Better than Graph

Level 1 val3 val5
Level 2 val1 val2 val4
?P(R)val3,val5
7
Questions

• Q1
• Is this model better than the model
  introducedby Agrawal and Wimmer?

8
Questions

• Q2
• How do we use top-k with this model?