An Agent-based Approach for Allocation between Buyers and Sellers through a Broker Agent in E-commerce Systems
67
Le Dien Tuan and Phan Dang My Phuong
An Agent-based Approach for Allocation between Buyers and Sellers
through a Broker Agent in E-commerce Systems
Le Dien Tuan1 and Phan Dang My Phuong2
1 Faculty of E-commerce, University of Economics, The University of Danang, Vietnam
2 Faculty of Banking, University of Economics, The University of Danang, Vietnam
Abstract. A broker agent in e-commerce systems enables buyers and sellers to do business with each
other to achieve the benefits for both buyers and sellers. A broker agent acts as a middleman between
buyers and sellers to carry out the allocation process. Although some recent research has focused on
this issue, an agent-based e-commerce system to carry out allocation buyers’ requirements and sellers’
offers is still sparse. Based on the mentioned analysis, this paper proposes an agent-based approach for
allocation between buyers and sellers through a broker agent in e-commerce systems. The major con-
tributions of this paper are that (i) an agent-based framework is proposed for allocating buyers and
sellers. The framework includes three layers: interface layers, modeling buyer’s requirements and
seller’s offers, and an allocation layer; (ii) a formula system is built to determine buyers’ satisfaction
degree as per seller’s offers; and (iii) an objective function is built to help a broker agent to carry out
allocation processes between buyers and sellers to maximize buyers’ satisfaction degree. The results of
simulation experiment demonstrate the proposed approach is effective under the different situations, it
provides a set of solutions to achieve optimal matching pairs to support decision-makers.
Keywords: e-commerce systems, an objective function, a broker agent, buyers’ satisfaction degree
1
Introduction
In the recent years, the internet becomes more popular in the world so e-commerce has developed digital
business environments and can replace more and more from the conventional business [1]. Furthermore,
with the development of information technology infrastructure and internet, more and more enterprises
carry out business transactions through internet so both buyers and sellers face with abundant information
from digital business environments. It is difficult for them to filter useful information to make a decision.
Thus, a broker agent emerges to solve the challenge. In particular, an electronic broker can gather useful
information of buyers and sellers and propose efficient ways to find a suitable trading partner [2-3].
At present, research on brokers or intermediaries as the third party of the trading processes in e-com-
merce has developed in recent years [4-5]. Sim et al. [6] proposed the protocol of broker to allocate buyers
and sellers under considering multiple attributes. Srivastava et al. [7] studied modeling and managing at-
tributes in a seller's offers through a broker to select the best seller as per buyers' requirements. Li et al. [8]
proposed a new approach to match buyers and sellers in a B2B e-marketplace through a matchmaker by
using multi-objective optimization.
Although the mentioned approaches above have focused on studying brokers as the third party in the
trading process between buyers’ requirements and sellers’ offers, there is little theory and few guidelines
related to allocation process based multi-agent systems. The lack of agent-based optimization allocation
approaches could not provide a solid foundation for improving exchange and market efficiency under the
consideration of buyers or sellers. Therefore, the novel approach in this paper is proposed to solve the
challenge.
The rest of this paper is organized as follows. Description of intelligent agent and multi-agent systems
in e-commerce is presented in Section 2. The proposed agent-based approach is introduced in Section 3.
An experiment is presented in Section 4. Section 5 concludes and points out our future work.
68
KỶ YẾU HỘI THẢO KHOA HỌC QUỐC GIA CITA 2020 “CNTT VÀ ỨNG DỤNG TRONG CÁC LĨNH VỰC”
2
Intelligent Agent and Multi-Agent Systems in E-commerce Application
E-commerce are becoming more and more dominant in modern transaction environments in recent years
because of their preeminence in automatic electronic transactions and low cost. On the other hand, agent
technologies are one of the most useful and powerful technologies to make e-commerce more distinct.
Thus, agent technologies and e-commerce are among the most important and exciting areas of research and
application development in e-commerce platforms. Combining the two fields offers important opportunities
for organizations to carry out automatic electronic transactions based on web environments and for devel-
opers to develop tools to facilitate business transactions in e-commerce [9]. Furthermore, intelligent agent
and multi-agent systems in e-commerce are a promising tool to solve commercial activities such as infor-
mation gathering, auctions, negotiation, shopping, trading and so on [10].
An agent is an intelligent software and it can help users to carry out communication each other [11].
Agents can carry out major actions in electronic transaction systems. Firstly, they are able to act on behalf
of their owners to find useful information and fulfil electronic transactions in e-commerce. Secondly, they
can locate potential partners to satisfy buyer’s requirements with less cost. Thirdly, they can make rational
decisions on behalf of humans and negotiate the price of a commodity with other agents to make a deal.
Agents in market-based environments can carry out functions such as buying and selling products and
other services. Other key applications of agents in e-commerce including negotiation, matching, locating
potential partners, managing supply chain, auctions, coalition formation and so on [12]. Furthermore, it is
expected that partially of fully automating some of the transaction processes in market-based multi-agent
environments will achieve significant cost savings. Agents act on behalf of buyers, sellers, brokers, vendors,
manufacturers, etc. in order to achieve business goals. To make market-based multi-agent environments
successful, agents should be able to carry out tasks in intelligent ways such as analysis, decision-making
and negotiation thorough reasoning and learning [13].
The objective of this paper is to work out the general model of e-commerce platforms targeted at alloca-
tions using the implementation of intelligent agent and multi-agent systems. In particular, we propose the
novel approach for allocation between buyers and sellers through a broker agent under the consideration of
multi-attribute trading in e-commerce systems.
3
Agent-based Approach for Allocation between Buyers and Sellers through a
Broker Agent
The proposed approach for solving the allocation problem between buyers and sellers through a broker
agent under the consideration of multi-attribute trading is presented in Figure 1.
In the framework, there are three layers for allocation process: the interface layer, modeling buyers’
requirements and sellers’ requirements, and allocation layer.
69
Le Dien Tuan and Phan Dang My Phuong
Buyer
Seller
Internet
Internet
Buyer agent
Seller agent
Submitting
Submitting
Broker agent
Evaluating
Evaluating
Allocation between buyers and sellers
Information process
Allocation model
Maintenance
The proposed formula system to calculate buyer’s satisfaction degree
Buyers’ requirements Sellers’ requirements
An Agent-based approach for allocation process
Fig 1. The framework of allocation process through a broker agent
Interface layer
3.1
The users, i.e., buyers, sellers can communicate with the system through menu-driven dialogue to express
their requirements. For example, a buyer starts the simplified interactive procedure with the system to help
the system to model buyers’ satisfaction degree as per a certain attribute. The interface layer helps buyers
and sellers to provide trading information to carry out the allocation process through a broker. The trading
information of buyers and sellers is related to carry out registration as well as provide other requirements.
On behalf of a buyer, a buyer agent can use its owner’s information, goods requirements with multi-attrib-
utes, and so on to provide them a broker agent. Furthermore, it can make a decision when its request has
been provided from a broker agent. A buyer agent’s functions consist of submitting and evaluating. A buyer
agent’s specific functions are explained as follows.
· Submitting: buyer’s requirements for certain products or services with multi-attributes are submitted to
the e-commerce systems by buyer agents.
· Evaluating: A buyer agent evaluates each sellers, which a broker agent sends to a buyer agent as per
buyer’s mentioned requirements.
A seller agent’s functions are similar as a buyer agent’s functions such as submitting and evaluating but
a seller agent acts on behalf of a seller.
3.2
Calculating buyers’ satisfaction degree based on sellers’ offers
Let Si be a set of seller {S1, S2,…,Sm} which is qualified to match ꢀꢁ, Sit denotes a set of values for attribute
(
)
At in Si (t∈ h+k ), ℎ is a number of attributes with hard constraints, ꢂ is a number of attributes with soft
constraints,
Let a set of buyers ꢀ = {ꢀꢃ, ꢀꢄ, … ꢀꢅ}, a set of sellers ꢆ = {ꢆꢃ, ꢆꢄ, … ꢆꢇ}, buyer’s requirements and
seller’s offers are related to multi attributes which can be split into attributes with hard constraints and soft
constraints. It is convenient to present the calculation method of buyer’s satisfaction degree, group of at-
tributes in buyer’s requirements is divided into a set of attributes with hard constraints ꢈ = {ℎꢃ, ℎꢄ, … ℎꢉ}
and a set of other attributes ꢊ = {ꢋꢃ, ꢋꢄ, … ꢋꢌ} [8].
70
KỶ YẾU HỘI THẢO KHOA HỌC QUỐC GIA CITA 2020 “CNTT VÀ ỨNG DỤNG TRONG CÁC LĨNH VỰC”
Seller’s offers are provided to a broker agent and these offers contains the same kinds of attributes re-
ꢔ
ꢒ
quested by buyers. Let ꢍ be number of sellers, ꢎ ∈ [1, … , ꢍ] is the index of ꢏ ’s offers and ꢑ (ꢓ ∈ ℎ +
ꢐ
ꢐꢁ
ꢂ) is the crisp value of the ꢓꢔꢕꢉ attribute of ꢆ ’s offers.
ꢐ
ꢔ
ꢒ
Let ꢖ be number of buyers, ꢓ ∈ [1, … , ꢖ] is the index of ꢀꢁ’s requirements and ꢗꢁꢁ (ꢓ ∈ ℎ + ꢂ) is the
crisp value of the ꢓꢔꢕꢉ attribute of ꢀꢁ’s requirements.
t
[
]
Based on notation of buyer’s requirements and seller’s offers above, gij∈ -1,1 is defined as a buyer's
satisfaction degree for the Atth attribute between ꢀꢁ and ꢆ . In particular, a buyer's satisfaction degree is
ꢐ
'
computed to attributes with hard constraints and soft constraints called gigj (g'∈h) and gigj(g∈k), respectively
as follows:
(i) For an attribute type with hard constraints:
-1 if C ≠Q
'
'
'
ig
jg
gigj = ꢘ
(1)
1
if C =Q
ig
jg
'
'
'
'
gg =-1 means that a seller ꢆ does not match with a buyer ꢀꢁ for attribute g' and gg =1 means that a seller
ꢐ
ij
ij
ꢆ matches with a buyer ꢀꢁ for attribute g'.
ꢐ
(ii) For an attribute type with benefit soft constraints: if cig>Q then ggij=-1. It means that a seller ꢙꢚ
jg
does not satisfied a buyer ꢛꢜ. If Cig≤Q , then ggij is calculated as follows:
jg
t'
Q
-Q
+∅
ggij=( Q
) ,
jg min-g
-Q
+∅
(2)
max-g min-g
where t'=Cig/Qmin-g, Qmin-g is the minimal value of a seller in the set of values for the attribute Ag and Qmax-g
is the maximal value of a seller in the set of values for the attribute Ag. A value t'∈(0,1] helps a broker to
carry out comparing a buyer’s satisfaction degree when t' is used to calculate ggij. ∅= Qmin-g/2, ∅ helps a
broker to solve some special cases such as only one seller in an e-marketplace or Qmax-g=Qmin-g. ggij increases
when Qig increases or Cig decreases.
ggij means that a seller ꢆ matches with a buyer ꢀꢁ for attribute g with a buyer’s satisfaction degree
ꢐ
'
t
Q
-Q
+∅
(
) . ggij is in-between 0 and 1. If gigj is near 1, it means that ꢀꢁ is highly satisfied by ꢆ for
jg min-g
ꢐ
Q
-Q
+∅
max-g min-g
attribute g.
(iii) For an attribute type with cost soft constraints: if Cig<Q then ggij=-1. It means that a seller ꢆ does
ꢐ
jg
not satisfied a buyer ꢀꢁ. If Cig≥Q , then ggij is calculated as follows:
jg
1
Q
-Q +∅
max-g jg
ggij=( Q
)t',
(3)
-Q
max-g min-g
+∅
ggij means that a seller ꢙꢚ matches with a buyer ꢛꢜ for attribute g with a buyer’s satisfaction degree
1
Q
-Q +∅
)t'.
max-g jg
(
Q
-Q +∅
max-g min-g
ggij is in-between 0 and 1. If ggij is near 1, it means that ꢛꢜ is highly satisfied by ꢙꢚ for attribute g. gigj in
this case increases when Qjg deceases or Cig increases.
(iv) For an attribute type with benefit interval constraints:
-1
if Qjg<CigL
(4)
Q − CigL
jg
if CigL≤Qjg<CigU
if Qjg≥CigL
ggij=
CigU − CigL
ꢝ
1
(v) For an attribute type with cost interval constraints
71
Le Dien Tuan and Phan Dang My Phuong
-1
if Qjg>CigL
(5)
CjgU − Q
jg
if CigL≤Qjg<CigU
if Qjg≤CigL
ggij=
CigU − CigL
ꢝ
1
In summary, a broker considers ꢛꢜ’s satisfaction degree based on ꢙꢚ's offers under multi-attribute
exchanges. Attributes with hard constraints are necessary conditions in trading processes and must be sat-
isfied. Thus, attributes with hard constraints do not need their weight. If attributes with hard constraints are
not satisfied, then ꢛꢜ cannot match with ꢙꢚ. On the other hand, attributes with soft constraints are necessary
for using the weight because these attributes can be relaxed within the given scope of values. In particular,
ꢛꢜ’s total satisfaction degree based on ꢙꢚ’s offers related to all attributes are calculated as follows.
wigggij ,
k
g=1
(6)
∑
k
∑
where wig is a weight value of attribute Ag for ꢀꢜ’s requirements, g=1 wig =1. In this paper, each ꢀꢜ
expresses the complete weight information for attributes in ꢀꢜ’s requirements.
3.3
Building allocation model between buyers and sellers
The broker agent’s allocation process is related to model information process of buyer’s requirements and
seller’s offers. In particular, modeling the information process as per multi-attribute to calculate buyer’s
satisfaction degree as per seller’s offers is discussed in subsection 3.2 above. This subsection is to build
allocation model to help broker agent to carry out allocations. Broker agent’s allocation process can also
be considered based on the optimization model because the goal of allocation between buyers and sellers
is to find a set of optimal pairs to maximize buyer’s satisfaction degree based seller’s offers. Based on a
broker agent’s responsibility, an objective optimization function and a set of constraints is built as follows.
ꢣ∈ꢌ ꢢ ggꢦꢁꢐ )
(7)
∑
∑
∑
ꢡ =
(
ꢁ∈ꢥ ꢐ∈ꢤ
ꢁꢣ
ij
∑
s.t. ꢁ∈ꢥ ꢦꢁꢐ ≤ 1, i = 1, 2, … n; j = 1, 2, … m
(8)
(9)
∑
ꢐ∈ꢧ ꢦꢁꢐ ≤ 1
ꢦꢁꢐ = 1 or ꢦꢁꢐ = 0
(10)
(11)
∑
ꢣ∈ꢌ wig = 1
ꢦꢁꢐ = 0 ꢓꢡ ggij =-1, gigj'=-1
or C ≠Q
(12)
'
'
ig
jg
Objective function in Equation 7 is to maximize the weighted sum of buyer’s satisfaction degree as per
seller’s offers; Constraints in Equation 8 and 9 are that each buyer (seller) can buy (sell) one unit of com-
modity at most. Constraint in Equation 10 are decision variable constraints, if ꢀꢁ matches with seller ꢆ ,
ꢐ
then ꢦꢁꢐ = 1; otherwise ꢦꢁꢐ = 0. Constraint in Equation 11 denotes the weight information of each buyer,
and constraint in Equation 12 determines constraint satisfaction layer to attend broker agent’s allocation
process. Furthermore, an objective optimization function in equation 7 can be efficiently solved by well-
known linear programming methods, such as simplex methods or interior point method [14].
A broker agent’s maintain strategy is carried out through updating information transactions of sellers
and buyers. When a broker agent receives new information transaction such as appearance of new buyers
or sellers, the broker agent’s allocation process is carried out again to find a set of optimal pairs.
72
KỶ YẾU HỘI THẢO KHOA HỌC QUỐC GIA CITA 2020 “CNTT VÀ ỨNG DỤNG TRONG CÁC LĨNH VỰC”
4
Experiments
In this section, we present our experimental results and analyse our allocation approach’s performance.
The experiments focus mainly on testing maximizing buyers’ satisfaction degree based on sellers’ offers
with multi-attributes in business environments. The rest of this section is divided into two subsections.
Selection 4.1 describe the experimental setting and selection 4.2 show the experimental results and perfor-
mance analysis in the different situations.
4.1
Experimental setting
In the experiments, we generate an artificial data of 10 buyer agents as per 10 buyer’s requirements and 20
seller agents as per 20 seller’s offers related to buy and sell the house in Australia. Buyer’s requirements
and seller’s offers related to house contains 5 attributes, i.e., price (a1), locations (a2), land size (a3), mov-
ing out time (a4), payment methods (a5). As per house buyer’s view, location attribute (a2) and payment
method attribute (a5) are attributes with hard constraints because their constraints must be satisfied while
price attribute (a1), land size attribute (a3), and moving out time attribute (a4) are attributes with soft
constraints because buyers can be accepted to relax the values of three attributes for their improved benefits.
In this experiment, the proposed allocation approach is evaluated under the change of different sellers in
business environments. More specifically, a broker agent’s allocation approach is tested in two different
experiments presented in Table 1 to maximize buyers’ satisfaction degree as per different sellers.
Table 1. Experiments based on different sellers
Experiments
1
Test purpose
The number of buyers is equal to the number of
sellers in business environments
2
The number of buyers is as twice as the number of
sellers in business environments
4.2
Experimental results and analysis
In experiment 1, a broker agent use the proposed approach to allocate house buyers and house sellers to
maximize house buyer's total satisfaction degree under considering that a number of house buyers (10
buyers) equals to a number of house sellers (10 sellers). In general principle of markets, when buyer's
demand equals to seller's supply, it is difficult for a broker agent to find out potential sellers to achieve their
high satisfaction degree because a broker has a fewer opportunity to choose seller's offers to satisfy buyer's
requirements. The results of buyers' satisfaction degree in experiment 1 are presented in Figure 2 and the
matching results are presented in Table 2.
Table 2. Optimal matching pairs with two different scenarios
Order
Experiment 1
Experiment 2
1
2
3
4
5
6
7
8
ꢀꢃ ↔ ꢆꢨ
ꢀꢄ ↔ ꢆꢩ
ꢀꢪ ↔ ꢆꢫ
ꢀꢬ ↔ ꢆꢄ
ꢀꢫ ↔ ꢆꢃ
ꢀꢨ ↔ ꢆꢭ
ꢀꢩ ↔ ꢆꢮ
ꢀꢭ ↔ ꢆꢃꢯ
ꢀꢮ ↔ ꢆꢪ
ꢀꢃꢯ ↔ ꢆꢬ
ꢡ = 0.8
ꢀꢃ ↔ ꢆꢃꢬ
ꢀꢄ ↔ ꢆꢫ
ꢀꢪ ↔ ꢆꢭ
ꢀꢬ ↔ ꢆꢃ
ꢀꢫ ↔ ꢆꢃꢩ
ꢀꢨ ↔ ꢆꢮ
ꢀꢩ ↔ ꢆꢃꢨ
ꢀꢭ ↔ ꢆꢄ
ꢀꢮ ↔ ꢆꢃꢯ
ꢀꢃꢯ ↔ ꢆꢃꢪ
ꢡ = 0.91
9
10
73
Le Dien Tuan and Phan Dang My Phuong
Based on Figure 2 and Table 2, it is clear that there are ten buyers are satisfied based on a broker agent’s
allocation process. However, each buyer’s satisfaction degree is not high. In particular, buyer’s minimal
satisfaction degree is 0.68 and buyer’s maximal satisfaction degree is 0.89. Furthermore, ten buyer’s nor-
malized total satisfaction degree in experiment 1 is not high (0.8) because a number of sellers equals to a
number of buyers.
Similarly, in experiment 2, a broker considers that a number of sellers (20) is twice as equal as a number
of buyers (10). Based on Figure 2 and Table 2, it is clear that the allocation results describe in Figure 2 and
Table 2. In particular, buyer's minimal satisfaction degree is 0.85 and buyer's maximal satisfaction degree
is 0.96. Furthermore, buyer's normalized total satisfaction degree in experiment 2 is relative high (0.91) and
is higher than buyer's normalized total satisfaction degree (0.8) in experiment 1 because a broker has many
opportunities to select potential sellers which satisfy buyer's requirements and increase buyers' total satis-
faction degree.
In summary, the proposed allocation approach perfectly works under different situations in business
environments. In general, if a number of sellers is more than a number of buyers, a broker agent has many
opportunities to choose seller’s offers to satisfy buyer's requirements and increase each buyer's satisfaction
degree as well as buyers’ total satisfaction degree.
Fig. 2. House buyer’s satisfaction degree
5
Conclusion and future work
The agent-based approach in this paper is proposed to solve allocation between buyers and sellers through
a broker agent. The proposed approach is novel because (i) the proposed framework includes three layers
via a broker agent, i.e., interface layer, modeling buyer’s requirements and seller’s offers and allocation
layer. Furthermore, functions of buyer agents and seller agents are clearly defined; (ii) a formula system is
proposed to calculate buyer’s satisfaction degree under the consideration of multi-attributes; and (iii) to
allocate buyers and sellers effectively, the objective functions and a set of constraints is built. The experi-
mental results demonstrate the good performance in the proposed approach in aspects of satisfying a buyer's
requirements and maximizing buyers’ total satisfaction degree.
Future research includes extending the proposed approach to solve competition environments and nego-
tiation between buyers and a broker in web-based environments.
74
KỶ YẾU HỘI THẢO KHOA HỌC QUỐC GIA CITA 2020 “CNTT VÀ ỨNG DỤNG TRONG CÁC LĨNH VỰC”
References
1.
Alpár, F. Z.: Matchmaking framework for b2b e-marketplaces. Informatica Economica, 2010.
14(4): 164-169, 2010.
2. Ketter, W., Collins, J., Reddy, P. & Weerdt, M.: The 2012 power trading agent competition. ERIM
Report Series Reference No. ERS-2012-010-LIS, pp. 1-49, 2012.
3.
Kuate, R. T., He, M., Chli, M. & Wang, H.: An Intelligent broker agent for energy trading: An
MDP approach. In: the Twenty-Third International Joint Conference on Artificial Intelligence, pp.
234-240, 2013.
4. Wu, L., Garg, S.K., Buyya, R., Chen, C. & Versteeg, S.: Automated SLA negotiation framework
for cloud computing. In: the 13th IEEE/ACM International Symposium on Cluster, Cloud and
Grid Computing (CCGrid), pp. 235-244, 2013.
5.
Wang, X., Zhang, M. & Ren, F. and Ito, T.: GongBroker: a broker model for power trading in
smart grid markets. In: the IEEE/WIC/ACM International Conference on Web Intelligence and
Intelligent Agent Technology (WI-IAT), pp. 21-24, 2015.
6. Sim, K. M., Chan, R.: A brokering protocol for agent-based e-commerce. IEEE Transactions on
System, Man, and Cybernetics Part C: Application and Reviews, Vol. 3, pp. 474-484, 2000.
7. Srivastava, N. K., Singh, P. & Singh, S.: Optimal adaptive CSP scheduling on basis of priority of
specific service using cloud broker. In : the 9th International Conference on Industrial and
Information Systems, pp. 1-6, 2014.
8. Li, X. & Murata, T.: Priority based matchmaking method of buyers and suppliers in b2b e-
marketplace using multi-objective optimization. In: the international MultiConference of
Engineers and Computer Scientists, Vol. 1, 2009.
9. K. M. Tolle and H. Chen.: Intelligent software agents for electronic commerce. Handbook on
Electronic Commerce, pp. 365–382, 2000.
10. M. J. North, C. M. Macal, J. St. Aubin, P. Thimmapuram, M. Bragen, J. Hahn, J. Karr, N. Brigham,
M. E. Lacy, and D. Hampton.: Multiscale agent-based consumer market modeling. Complexity,
15(5):37–47, 2010.
11. Jung, J.-J., Jo, G. S.: Brokerage between buyer and seller agents using constraint satisfaction
problem models. Decision Support Systems, 28(4): 293-304, 2000
12. A. Negri, A. Poggi, M. Tomaiuolo, and P. Turci.: Agents for e-business applications. In
Proceedings of the Fifth International Joint Conference on Autonomous Agents and Multiagent
Systems, pp. 907–914, 2006.
13. R. Rajavel and M. Thangarathanam.: Adaptive probabilistic behavioural learning system for the
effective behavioural decision in cloud trading negotiation market. Future Generation Computer
Systems, 58:29–41, 2016
14. Fletcher, R.: Practical methods of optimization. John Wiley & Sons, 2013.
8 trang
10000
Free
Bạn đang xem tài liệu "An Agent-based Approach for Allocation between Buyers and Sellers through a Broker Agent in E-commerce Systems", để tải tài liệu gốc về máy hãy click vào nút Download ở trên
File đính kèm:
- an_agent_based_approach_for_allocation_between_buyers_and_se.pdf
