Wednesday June 28, 2000
2:00 - 4:30 p.m.
The Bank of Nova Scotia,
This is 1 block north of King at the south-east corner of Yonge and Adelaide (King subway station).
The development of a robust client segmentation solution is a key prerequisite to the design of strategic new value propositions in financial services. However, the computationally intensive nature of the underlying clustering techniques currently being employed to build these systems requires that the analyst make tradeoffs between development time and the effectiveness (robustness and relevance) of the final solution. The analyst must often rely heavily on the "business experience" of the system end-users to help sort through the massive number of potential system inputs, such as the number of clusters, number of driver variables and model inputs. Experience has shown that different analysts will come up with very different solutions, often to the frustration of the end users.
We have addressed these issues by developing an objective function which measures both the relevance of the final solution and the "distinctness" of the clusters themselves. This objective function is then used by a genetic algorithm (GA) to evaluate, rank, and evolve progressively better solutions using the simple operators of selection, crossover, and mutation. The GA enables us to do a constrained (# of cluster, # of inputs) yet highly parallel search through all possible combinations of cluster driver variables, and leads to closer-to-optimal solution in much less time than previously possible. Our talk will focus on the development of the objective function, and the challenges we faced encoding the problem.
The application of genetic algorithms to segmentation problems.
A walk-through of our Excel/SAS application.
Senior Manager Market Segmentation, Bank of Montreal.
B.Eng, Industrial Engineering (Ryerson), MBA, Information Systems (McMaster)
Senior Research Analyst Market Segmentation, Bank of Montreal.
B.Eng, Mechanical Engineering (McMaster), MBA Finance (McMaster)
Due to limited space, we require all attendees to preregister with Tamara Arenovich, Vice-Chair, SORA Committee for Financial Statistics & Marketing:
Centre for Academic Technology
University of Toronto
Phone: (416) 978-5128
Fax: (416) 978-7705
For those who register colleagues and guests for this seminar, please make sure that you include the name, phone number, company name/institution name, e-mail address and fax number of those people so that we will have a proper record in our system.
For more information about this seminar and other SORA events and activities, please contact either Tamara or Ivan Hon, Chair, SORA Committee for Financial Statistics & Marketing:
Senior Technical Manager
The Bank of Nova Scotia
602-100 Yonge Street,
Toronto, Ontario M5H 1H1
Phone: (416) 933-1493