Page 278 - Data Science Algorithms in a Week

Agent-Based Modeling Simulation and Its Application to Ecommerce 259

System Dynamics

The utility of system dynamics (SD) modeling approach is well documented in
literature. SD is a non-data driven system thinking approach that targets top management.
This is convenient since detailed data or business process activities are not always
available. SD is a continuous simulation methodology whose models are more intuitive
than discrete-event simulation. This methodology lends its expertise to dynamic problems
of strategic importance for varying horizons.
The interest of SD is not in the implementation of individual events but in aggregate
terms. Several studies adopt SD to model overall structure of the organization at strategic
and tactical management levels as well as to capture financial and global environment
(Rabelo, et al. (2007); Rabelo et al. (2005)).
Speller et al. (2007) developed a system dynamics model to capture dynamic value
chain system of “the traditional production/assembly supply chain with service
components added to it.” The first step is made up of generic, causal-loop diagrams and
subsequently a detailed stock-and-flow model. Taylor series approximations were used to
generate a linear system of differential equations to capture the behavior of the system
with time. These behaviors are analyzed and make long-range predictions of interest
using the eigenvalue technique. SD serves as a response to the inadequacy of the
application of operation research and other management science methodologies for
solving complex problems with large number of variables, nonlinearity and human
intervention.
SD modeling captures physical laws governing a system using subjective thinking
with an assumption of dynamic behavior of entities (An and Jeng, 2005). Due to
complexity characterized by nonlinearity and time delay, the system may not be solved
analytically. Available numerical method for ordinary differential equations such as
Euler’s first order finite difference, Runge-Kutta second and fourth order finite difference
method can be employed to solve the system numerically.
System dynamics models have been used to represent and analyze different aspects
of the e-commerce business. Causal loops diagrams are useful to capture the structure of
e-business systems (Kiani, Gholamian, Hamzehei, & Hosseini, 2009) and to understand
how positive and negative feedbacks have impact on the strategies designed for online
markets (Fang, 2003; Oliva, Sterman, & Giese, 2003).
Topics of study using SD in the internet environment, such as consumer behavior
(Khatoon, Bhatti, Tabassum, Rida, & Alam, 2016; Sheng & Wong, 2012) and credit risk
analysis (Qiang, Hui, & Xiao-dong, 2013) are examples of important aspects considered
when modeling online trading. System dynamics models are widely used as policy
laboratories to find the appropriate strategies to reduce cost and increase revenue. This
type of research has also been applied to the online marketplace (Liping An, Du, & Tong,