incentive y. liu anincentivemechanismdesignedfor e ... · incentive mechanism y. liu introduction...
TRANSCRIPT
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
An Incentive Mechanism Designed forE-Marketplaces with Limited Inventory
Yuan Liu, Jie zhang
School of Computer Engineering
Nanyang Technological University
August 03, 2013
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 1 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Overview
1 Background and motivation
2 Our incentive mechanism
3 System analysis
4 Initialization
5 Experimental validation
6 Conclusions
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 2 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Introduction
Reputation system
Modeling sellers’ honesty based on buyers’ ratings
Problems:
Unfair rating problemRe-entry problem
Existing Mechanisms to Address the Unfair Rating Problem
Side-payment mechanism [Jurca et.al]
Reward maximal side payment for truthful ratings
Trust-based approach [Zhang et.al]
Honest buyers can propagate seller reputation moreeffectively, then be offered with lower prices
Trust-revelation [Sviatoslav et.al]
Truthfully revelation seller trustworthiness can bring sellersmaximal utility
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 3 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
E-marketplaces with Limited Inventory
One common assumption: there is unlimited products providedby sellers, i.e. no competition between buyers
E-marketplaces with Limited Inventory (EMLI)
Definition: Given a set of sellers S each of whom providesthe same product and a set of buyers B each of whomdemands one piece of the products. An e-marketplacesatisfying the condition, |S| < |B|, is called ane-marketplace with limited inventory.
Examples:
Second-hand textbooksHotel booking in peak seasonsDoctor booking system
New challenges in promoting buyer and seller honesty
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 4 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
E-marketplaces with Limited Inventory
One common assumption: there is unlimited products providedby sellers, i.e. no competition between buyers
E-marketplaces with Limited Inventory (EMLI)
Definition: Given a set of sellers S each of whom providesthe same product and a set of buyers B each of whomdemands one piece of the products. An e-marketplacesatisfying the condition, |S| < |B|, is called ane-marketplace with limited inventory.
Examples:
Second-hand textbooksHotel booking in peak seasonsDoctor booking system
New challenges in promoting buyer and seller honesty
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 4 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
E-marketplaces with Limited Inventory
One common assumption: there is unlimited products providedby sellers, i.e. no competition between buyers
E-marketplaces with Limited Inventory (EMLI)
Definition: Given a set of sellers S each of whom providesthe same product and a set of buyers B each of whomdemands one piece of the products. An e-marketplacesatisfying the condition, |S| < |B|, is called ane-marketplace with limited inventory.
Examples:
Second-hand textbooksHotel booking in peak seasonsDoctor booking system
New challenges in promoting buyer and seller honesty
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 4 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
New Challenges
Buyers have incentives to be dishonest
‘positive’ rating for a good seller בnegative’ rating for a good seller
√
‘negative’ rating for a bad seller בpositive’ rating for a bad seller
√
Sellers have incentives to be dishonest
Honest delivery → increased cost ×Dishonest delivery → decreased cost + sold out
√
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 5 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
New Challenges
Buyers have incentives to be dishonest
‘positive’ rating for a good seller בnegative’ rating for a good seller
√
‘negative’ rating for a bad seller בpositive’ rating for a bad seller
√
Sellers have incentives to be dishonest
Honest delivery → increased cost ×Dishonest delivery → decreased cost + sold out
√
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 5 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
The Objective of This Paper
We have done
Design an incentivemechanism to promotebuyer and seller honestyfor the e-marketplaceswith limited inventory
Modeling buyer honestyModeling seller honestyPricing algorithmAllocation algorithm
Initialization the system
Initial honesty valuesMembership fees
Important Achievements
Buyers have incentives tobe honest
Sellers have incentives tobe honest
Buyers and sellers have noincentive to re-enter
Experimental validation
Static simulationDynamic SimulationRe-entry simulationComparison Results
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 6 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
The Objective of This Paper
We have done
Design an incentivemechanism to promotebuyer and seller honestyfor the e-marketplaceswith limited inventory
Modeling buyer honestyModeling seller honestyPricing algorithmAllocation algorithm
Initialization the system
Initial honesty valuesMembership fees
Important Achievements
Buyers have incentives tobe honest
Sellers have incentives tobe honest
Buyers and sellers have noincentive to re-enter
Experimental validation
Static simulationDynamic SimulationRe-entry simulationComparison Results
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 6 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Our Incentive Mechanism
The components of the incentive mechanism
Modeling
Buyer Honesty
Modeling
Seller Honesty
Allocation
Algorithm
Pricing
Algorithm
Buyer Honesty
Promoted
Seller Honesty
Promoted
SellerReputation
BuyerRatings
Transaction
BuyerHonesty
+
Buyer ScoreSeller
Reputation
Figure: The work flow of our incentive mechanismY. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 7 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
We analyze the system in three aspects:
Individual rationality
Incentive compatibility
Social welfare
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 8 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Individual Rationality
Buyer Individual Rationality
Prop 1: Truthful ratings can maximize buyer scores.
Prop 2: Buyer utility is positive from a transaction if Rs > R0.
Prop 3: The upper bound of the price for buyers is C
R0.
Seller Individual Rationality
Prop 4: Utility is positive when seller reputation Rs > R0.
Prop 5: The lower bound of the price for sellers is C
δ.
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 9 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Incentive Compatibility
Buyer Incentives
Prop 6: More utility can be gained from a transaction if Rs ishigher.
Prop 7: The utility in providing truthful ratings is no less thanproviding untruthful ratings.
Seller Incentives
Prop 8: Sellers have incentive to improve their honesty bydelivering promised products.
Both buyers and sellers have incentive to be honest.
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 10 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Social Welfare
Definition of Social Welfare
W(Rs, Rb) = (Us(Rs)+Ub(Rs))1
1−Rb
= Rs(Vsb −C)
1
1−Rb
,
(1)
Prop 9: The proposed incentive mechanism can increase thetotal social welfare as defined in Equation (1).
Prop 10: The social welfare of the proposed system is no lessthan that of the free-trading e-marketplace.
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 11 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Initialization
Rb(0): 0
Rs(0): δ
To avoid re-entry, we assign membership feeM = N0(M1 +M2) = N0C where M1 = (1−R0)C,M2 = R0C, and N0 is determined by the Chernoff BoundTheorem based on the error rate.
When sellers leave the system, Mr will be returned tosellers
Mr=M(Rs)=
N0(M1 +M2) Rs > δ,
(Rs−R0
δ−R0)2N0M1+N0M2 R0≤ Rs≤δ,
0 Rs < R0.
(2)
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 12 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Experimental Validation
Settings
R0 = 0.6, δ = 0.85
C = 1, V∗ = 2, V ∗ = 2.5
η = 0.1, α = 0.5
Honest seller: delivering promised productsDishonest seller: delivering 50% quality productsHonest buyer: ‘1’ for honest transaction; ‘0.5’ fordishonest transactionDishonest buyer: ‘0.5’ for honest transaction; ‘1’ forhonest transaction
Bootstrap our system: 80 honest buyers and 40 sellers for1000 transactions
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 13 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Experiments
Experiments
Static experiments:
another 320 buyersconducting 9000 transactions
Dynamic experiments:
5 new sellers and 50 new buyers join in each 100transactionsconducting 400 transactions
Reentry experiments:
simulate the profit of sellers with reentry
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 14 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Static Experimental Results
Seller’s Honesty and Seller’s Profit
0.5
0.6
0.7
0.8
0.9
1
0 4 8 12 16 20 24 28 32 36 40 0.75
0.8
0.85
0.9
0.95
1
Selle
r H
ones
ty
Selle
r R
eput
atio
nSeller ID
(a)
HonestyReputation
0.5
0.6
0.7
0.8
0.9
1
0 4 8 12 16 20 24 28 32 36 40 0.1
0.15
0.2
0.25
0.3
Selle
r H
ones
ty
Selle
r Pr
ofit
Seller ID
(b)
HonestyProfit
Figure: The relationship between probability of sellers in behavinghonestly and (a) seller reputation, (b) average seller profit in thestatic setting
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 15 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Static Experimental Results (2)
Buyer’s Honesty and Buyer’s Utility
0.5
0.6
0.7
0.8
0.9
1
0 40 80 120 160 200 240 280 320 360 400 0.95
0.96
0.97
0.98
0.99
1
Buy
er H
ones
ty
Buy
er S
core
Buyer ID
(a)
HonestyScore
0.5
0.6
0.7
0.8
0.9
1
0 40 80 120 160 200 240 280 320 360 400 0⋅100
1⋅103
2⋅103
3⋅103
4⋅103
5⋅103
Buy
er H
ones
ty
Buy
er U
tility
Buyer ID
(b)
HonestyUtility
Figure: The relationship between probability of buyers in behavinghonestly and (a) buyer scores, (b) buyer total utility in the staticsetting
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 16 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Dynamic Experimental Results
Seller’s Honesty and Seller’s Profit
0.5
0.6
0.7
0.8
0.9
1
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0.75
0.8
0.85
0.9
0.95
1
Selle
r H
ones
ty
Selle
r R
eput
atio
nSeller ID
(a)
HonestyReputation
0.5
0.6
0.7
0.8
0.9
1
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0.1
0.15
0.2
0.25
0.3
Selle
r H
ones
ty
Selle
r Pr
ofit
Seller ID
(b)
HonestyProfit
Figure: The relationship between probability of sellers in behavinghonestly and (a) seller reputation, (b) average seller profit in thedynamic setting
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 17 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Dynamic Experimental Results (2)
Buyer’s Honesty and Buyer’s Utility
0.5
0.6
0.7
0.8
0.9
1
0 60 120 180 240 300 360 420 480 540 600 0.93
0.94
0.95
0.96
0.97
0.98
0.99
1
Buy
er H
ones
ty
Buy
er S
core
Buyer ID
(a)
HonestyScore
0.5
0.6
0.7
0.8
0.9
1
0 60 120 180 240 300 360 420 480 540 600 0⋅100
1⋅103
2⋅103
3⋅103
4⋅103
5⋅103
6⋅103
Buy
er H
ones
ty
Buy
er U
tility
Buyer ID
(b)
HonestyUtility
Figure: The relationship between probability of buyers in behavinghonestly and (a) buyer score, (b) buyer total utility in the dynamicsetting
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 18 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Reentry Experimental Results
Seller Reentry
-2
0
2
4
6
8
10
12
0 4 8 12 16 20 24 28 32 36 40
Tota
l Pro
fit o
f Sel
lers
No. of Period
(a)Rep=1
Rep=0.7Rep=0.55
-2
0
2
4
6
8
10
12
0 4 8 12 16 20 24 28 32 36 40
Tota
l Pro
fit o
f Sel
lers
No. of Period
(b)
ReentryPoint
Rep=1Rep=0.7
Rep=0.55
-4
-2
0
2
4
6
8
10
12
0 4 8 12 16 20 24 28 32 36 40
Tota
l Pro
fit o
f Sel
lers
No. of Period
(c)
ReentryPoint
Rep=1Rep=0.7
Rep=0.55
-2
-1.5
-1
-0.5
0
0.5
0.5 0.6 0.7 0.8 0.9 1
Prof
it Lo
ss o
f Sel
lers
Reputation
(d)
δ
Figure: Re-entry scenarios:(a) sellers’ total profit without re-entry and without membership fee,(b) sellers’ total profit with re-entry but without membership fee,(c) sellers’ total profit with re-entry and with membership fee,(d) profit loss of sellers by comparing (c) and (a)
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 19 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Comparison Results
Comparing Our Model with BRS in EMUL
0
1
2
3
4
5
6
4 8 12 16 20 24 28 32 36 40
Avg
Sel
ler
Prof
it
Seller ID
(a)
Honest Dishonest
Our MechanismSide-payment
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
40 80 120 160 200 240 280 320 360 400
Avg
Buy
er U
tility
Buyer ID
(b)
Honest Dishonest
Our MechanismSide-payment
Figure: The incentive comparison between our mechanism and theside-payment mechanism (a) seller incentive, (b) buyer incentive inEMUI
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 20 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Comparison Results (2)
Comparing Our Model with BRS in EMLI
0
0.2
0.4
0.6
0.8
1
4 8 12 16 20 24 28 32 36 40
Avg
Sel
ler
Prof
it
Seller ID
(a)
Honest Dishonest
Our MechanismSide-payment
0
0.1
0.2
0.3
40 80 120 160 200 240 280 320 360 400
Avg
Buy
er U
tility
Buyer ID
(b)
Honest Dishonest
Our MechanismSide-payment
Figure: The incentive comparison between our mechanism and theside-payment mechanism (a) seller incentive, (b) buyer incentive inthe EMLI
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 21 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
Conclusions
Conclusions
We proposed an incentive mechanism to
Promote buyer honestyPromote seller honesty
Theoretical analysis
Buyers have incentives to provide honest ratingsSellers have incentives to deliver promised productsThe social welfare is improved
Experimental analysis
Buyers gain more utility from honest ratingsSellers gain more profit from honest deliveryBoth buyers and sellers have no incentive to reenterThe proposed model perform better than BRS in bothEMLI and EMUL
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 22 of 23
Incentive
Mechanism
Y. Liu
Introduction
Model
Analysis
Initialization
Simulation
Summary
THANK YOU
Y. Liu, SCE, NTU BCSI2013 Incentive Mechanism page 23 of 23