Information Technology Reference
In-Depth Information
Table 6.
Absolute and relative error of the several configurations (payment methods:
Z
1
=invoice,
Z
2
=prepayment,
Z
3
=credit card,
Z
4
=cash on delivery; payment configu-
ration: 0=off, 1=on; error:
Δ
=absolute error,
δ
=relative error; direction:
↓
=simulation
value
<
reference value,
↑
=simulation value
>
reference value)
Z
1
Z
2
Z
3
Z
4
Cancel
Config.
Z
1
Z
2
Z
3
Z
4
Δδ
Δ δ
Δδ
Δδ
Δδ
1
0000
-
-
-
-
-
-
-
-
0
0%
3
0100
-
- 0.0314%
↑
-
-
-
-
0.03 4%
↓
6
1
1
0 0
0
0%
0
0%
-
-
-
-
0
0%
9
0110
-
- 0.0218%
↑
0.01 2%
↓
-
-
0.01 3%
↓
10
0
1
0 1
-
-
0
0%
-
-
0
0%
0
0%
13
1101 0.011%
↑
0.01 17%
↓
-
-
0
0%0
0%
16
1
1
1 1
0
0%
0
0%
0
0%
0
0%
0
0%
is the only alternative. If 77% of all consumers is female, then similar effects can
be observed.
4
Related Work
Multi-agent models are widely-used to simulate customer behavior. Moreover,
agent-based computational economics is a separate research field which uses
agent-based systems to study economies that are emerge from individual de-
cisions of autonomous agents and their interactions [10]. In the following, we
discuss four approaches for model consumer behavior.
A relevant contribution to our work has come from Rigopoulos et al [11]. The
authors describe a multi-agent system for the acceptance of payment systems.
The customer decision process of choosing a payment method is also simulated
using the utility theory. The consumer calculates the utility of each available
payment method resulting in a probability vector including an adopting prob-
ability for each payment method. In opposition to our model, the utility of a
payment method depends on several consumer-specific attributes. Another dif-
ference is the application area. Rigopoulos et al. focus on digital retail payments
in general, whereas our model addresses the payment process in online stores.
Since the model described in [11] is developed to support the strategic decisions
of banks and other payment service providers, it is also possible to forecast the
success of new payment methods. Our model considers the most widely-used
payment methods of the German market.
The agent-based simulation technique is also used to model the behavior of
customers in a supermarket [12]. In conformity with our approach, the authors
characterize every customer agent with specific attributes such as gender. Addi-
tionally, each consumer is characterized by a set of feature agents. Every feature
agent represents a single parameter of the consumer's behavior and is modeled as
an autonomous agent. The interaction of the feature agents yields the shopping
