CMPT 880: Topics in Computing Science:
Modeling of Complex Social Systems (Fall 2008)
Instructor : Vahid Dabbaghian (vdabbagh@sfu.ca)
Assistant Instructor : Warren Hare (whare@irmacs.sfu.ca)
Lecture room : 10940 (The IRMACS Centre)
Meeting times : Fridays 11:30 to 12:30
Web Page : http://www.sfu.ca/~vdabbagh/CMPT880.htm
DESCRIPTION OF THE COURSE:
This is a seminar course that reviews theory and research in complex social systems. In particular we will focus on the impact of social interactions on the dynamic of urban transformations such as crime and infectious diseases in municipal environments. The seminars incorporate mathematical modelling and computer simulations.
TOPICS:
Exact modeling techniques covered will vary with
class size and interest, but in general the following topics will be covered.
Good Modelling Practices: Simplicity, Adaptability, Reproducibility, Validation.
Complex Social Networks: What are they, why model them, examples.
Operational Management Models: System Dynamics, Scheduling, Queuing Models.
Forecasting Models: Regression Analysis, Markov Models, Discrete
Event Models.
Pattern Reconstruction Simulation Models: Cellular Automata, Network Models, Agent Based Models.
Announcements:
ASSIGNMENT
1 (due Sept 26th, in class):
Determine a
problem that contains a social network and could be approached via modelling.
Prepare a five minute talk outlining the problem. (pdf, ppt, or whiteboard are
acceptable)
Reading materials:
Meetings
Program:
Week 1:
Modelling Methodology: Good, Bad and Ugly
Week 2:
Speaker: Dr. Martin Andresen, Assistant Professor, School of
Criminology, Simon Fraser University
Title: Predicting local crime clusters using a local
indicator of spatial association and a discrete choice model .pdf
Abstract: The use of local spatial statistics in crime
analysis, particularly at the neighborhood level, is relatively sparse. Within
the research that does use local spatial statistics, the methods are used to
identify clusters of crime but not explain them. In this paper, local spatial
statistics are used to identify local clusters of crime that are then modeled
in a discrete regression to determine the correlates of local crime clusters. The
standard ecological theories of crime (social disorganization and routine
activity theory) perform relatively well in predicting local crime clusters,
but it is found surrounding neighborhoods are important in identifying the
predictors of different local crime cluster types.
Week 3:
Population and Crime Modeling (Katie Wuschke)
Facts
& Statistics (Azadeh Alimadad)
Week 4:
Speaker: Terry Howard, Coordinator, Prison
Outreach Program
Title: Prisoner Positive: Providing outreach support to
prisoners living with HIV/AIDS in BC. pdf
Abstract: This lecture will highlight the support services
provided to prisoners living with HIV/AIDS in BC, and discuss the impact that
support has on changing harmful behaviours and criminogenic factors in their lives. We will explore ways
the statistical data collected can be employed to improve service provision and
inform prison administrative policy for health promotion and harm reduction
programming in a time of budget cuts and “tough on crime” climate.
Week 5:
Dynamic
of urban transformations
Week 6:
Speaker: Dr. Bryan Kinney, Assistant Professor, School of
Criminology, Simon Fraser University
Title: Computational criminology and the possibility of
"best practices". pdf
Abstract: This talk will cover in brief what seems to be a
current focus on "evidence-based" policy development and evaluation
practices within the general area of crime prevention and crime reduction.
Chief among the issues to be discussed is the apparent rise of calls for
evidence-based policy to establish "best practices", and, of course,
its related concern, that being to find out "what works" in crime
reduction and crime prevention. While I argue that much has to be done to
improve our collective techniques to more accurately approach the validity and
the reliability that academics, police agencies--and especially government
officials--see as achievable in the name of best practices, the key to success,
as I see it, is the blending of the more traditionally 'social' sciences with
those that are more familiar with computational methods and analytic
strategies.
Week 7:
Week 8:
Speaker: Diane T. Finegood,
Scientific Director of Canadian Institutes of Health Research and Institute of
Nutrition, Metabolism and Diabetes.
Title: A complex systems perspective on obesity research,
policy and practice
Abstract: The global epidemic of obesity has emerged as a
result of many factors which influence food and
physical activity related behaviours. Increased
caloric intake, in particular through consumption of energy dense foods, and
decreases in daily levels of energy expenditure are associated with changes in
our social and physical environments. Changes in the social and physical
environment have occurred at multiple levels including those proximal to
individuals such as home, schools, and worksites, in communities and regions
and at national and international levels (IOTF Causal Web). Over the past few
decades overweight and obesity have usually been addressed with simple
solutions such as diets promoted by the diet industry or health promotion
campaigns encouraging individuals to increase their level of physical activity
or decrease consumption through means such as a low fat diet. These approaches
have generally ignored the complexity of the food and physical activity
environment and have not recognized the need to increase individual capacity or
decrease complexity. Recent attention to the epidemic of obesity has resulted in
an increase in collective effort and investment in research and programs aimed
at increasing physical activity and decreasing food intake. Although much of
this investment is still directed at education and social marketing campaigns,
there is some recognition that the complexity of the environment needs to be
reduced to make "the healthy choice the easy choice". Increased
investments in research and evaluation are also supporting the measurement of
effectiveness of policies and practice. Future efforts to apply a complex
systems analysis are needed to help target investments to solutions that will
increase capacity, decrease complexity, and enhance feedback loops that measure
the effectiveness of new policies and programs.
Week 10:
Speaker: Shane G. Henderson, Associate Professor in the School
of Operations Research and Information Engineering at Cornell University
Title: Ambulance Location and Relocation
Abstract: Ambulance service providers all over the world are
struggling to maintain service levels, as measured by response times to calls,
in the face of increasing traffic congestion, increasing call volumes, and budget
pressure. As a consequence, they are looking, more than ever, to provide a good
match between supply and demand of ambulances. The ambulance location problem
is that of determining how many ambulances to operate from each of a predefined
set of bases. We first review a simple model that sheds light on this question:
One should not allocate ambulances to bases in proportion to demand. The optimal
(static) allocation of ambulances to bases is important, but it ignores the
option of dynamically moving ambulances in real time to attempt to "fill
holes," which is also known as relocation. We discuss some existing
methods for relocation, as well as a new method based on approximate dynamic
programming (ADP). Computational results for Edmonton (Canada), and Melbourne
(Australia) show that ADP can work well, but a successful implementation
requires a mix of queueing-systems knowledge and intuition.
Week 12:
Speaker: Bojan Ramadanovic,
Complex Systems Modelling Group, Simon Fraser
University.
Title: Modelling surgical
wait-lists with queuing theory
Abstract: This talk deals with the segment of a surgical
wait-lists model for British Columbia, which is being developed by the Complex
Systems Modelling Group at the IRMACS centre.
Principal topic to be discussed will be the application of the number of
queuing theory results, such as queuing
with reneging, to the problem at hand. Talk will also address the interplay
between computational simulation and the analytical modelling.
