Information Technology Reference
In-Depth Information
decides on basis of a certain probability for an alternative payment method.
This decision essentially depends on the utility of a payment method which is
gender-neutral and described by the following function:
utility
:
Z →
R
.
(6)
In the literature, there are different approaches for modeling the consumer spe-
cific decision process using a utility for each alternative. We use the so-called
Luce model [4] which is defined as follows:
Z
i
)
i
=1
utility
utility
(
P
(
Z
i
)=
Z
i
)
.
(7)
(
Finally, if the customer finds one or more alternative types of payment, then
he will choose those alternatives with the specific probability and the purchase
process will end. In case she/he does not find an appropriate alternative, this will
lead to the abort of the purchase and the termination of the payment process.
2.3 Agent-Based Implementation of the Shopping Simulation
In the course of time, several simulation techniques have been developed. A rel-
atively new method is multi-agent simulation which has become a common ap-
proach for simulation in social science [5]. This technique uses intelligent agents
for the modeling of complex systems. Agents are entities in a certain environ-
ment with a limited perception of their environment. They are capable of doing
autonomous actions and interactions with other agents in order to delegate their
objectives [6].
Multi-agent systems are widely-used in the field of retailing. The heteroge-
neous behavior of the customers and their several interactions are well supported
by the agent-based paradigm. The underlying bottom-up approach allows for a
realistic modeling of consumer behavior. Furthermore, the approach enables the
usage of user-specific information collected from real systems such as Customer
Relationship Management (CRM) or e-commerce systems.
We apply the agent-based approach for modeling the heterogeneous payment
behavior depending on specific attributes. The implementation of a multi-agent
system is usually done by means of agent-based modeling platforms. Today, there
is a large number of such platforms available, see [7] or [8] for an overview of var-
ious toolkits. We use SeSAm (Shell for Simulated Agent Systems) to implement
the different parts of our shopping model. This open source tool supports a vi-
sual programming of the implementation. The behavior of the agents is specified
using an activity graph derived from UML [9].
The customers are modeled as agents. We specify the general class customer
and derive two special classes thereof. One represents male consumers, the other
characterizes the female consumers. The e-commerce system and its relevant
attributes are modeled as the environment of the agents. Resources are not used
within the model. The information exchange between the different parts of the
shopping simulation is realized via comma-separated files.
