Fuzzy logic, SD, and Business Policies

This forum contains all archives from the SD Mailing list (go to http://www.systemdynamics.org/forum/ for more information). This is here as a read-only resource, please post any SD related questions to the SD Discussion forum.
Locked
jnoble1@mmm.com
Newbie
Posts: 1
Joined: Fri Mar 29, 2002 3:39 am

Fuzzy logic, SD, and Business Policies

Post by jnoble1@mmm.com »

I would like to query the list on who is doing research in integrating
fuzzy logic and system dynamics for business management problems and
what's been recently published on the topic. A brief search of ""System
Dynamics Review"" for the past 5 years or so did not turn anything up. Here
is my motivation for asking the question. One of the pillars of SD is the
constant emphasis on transparency and ""understandability"" of the models
created for the clients and customers of a particular SD project. It seems
to me that fuzzy logic, with it's capability to quantify and execute
inferences with qualitative categories of phenomena, would be a beautiful
complement to traditional SD methods and goals. This would be especially
true around business policy issues, where I would think an interactive
formulation process with clients around capturing business rules using
qualitative words and in ""If-then"" relationships would be especially
helpful in making connections between abstract model constructs and actual
operating procedures, For example, a set of business rules in
manufacturing planning interrelating variables such as demand, order
back-log, and capacity rate changes could have appropriate states and
trajectories mapped to membership functions and then formulated as rules
such as the following: ""If demand is increasing rapidly and the order
back-log is large, a large increase in capacity is expedited"". A set of
such rules would seem to be 1.) much more intuitive than equations and
mathematical formulations and 2.) allow for nonlinear formulations as
typically occur in policies of any complexity. When trying to ""tune"" a
model with historical time series data, recent work in fuzzy modeling to
capture quantitative relationships with a minimum number of rules to
improve ""interpretability"" would seem to be very helpful for this
application. I just would like to know if anyone in the SD community has
been going down this path of research and what conclusions they have been
reaching.
Thanks,

Jonathan Noble
3M Supply Chain
ph. 641-828-5518
Tri: 828-5518
e-mail:
jnoble1@mmm.com
Joel Rahn
Junior Member
Posts: 15
Joined: Fri Mar 29, 2002 3:39 am

Fuzzy logic, SD, and Business Policies

Post by Joel Rahn »

A fuzzy relation between two variables can be de-fuzzified to make a
LOOKUP function Vensim (formerly TABLE function in DYNAMO) and a LOOKUP
function can represent a fuzzy relation between two variables. The whole
argument rests on a representation theorem for fuzzy relations that
says, essentially, a fuzzy relation can represent a piece-wise
approximation of an arbitrary function -i.e. an arbitrary LOOKUP
function. Going from a LOOKUP function to a useful fuzzy relation is
trickier -the membership functions seem to be constrained to be
triangular to make the fuzzification straightforward but the result may
not be intuitive; for example, there may be 'too many' triangular
membership functions as in Hot, Medium Hot, Warm, Medium Warm, Tepid,
Medium Tepid,..,..,..,..,Cold.

Joel Rahn
jrahn@sympatico.ca
Locked