Past Capstone Projects

A final term-long project sponsored by industry partners caps the certificate.  Information for organizations interested in participating can be found in the sponsor handbook. Some of the previous projects:

Passenger Railcar Maintenance

Data collected for strictly operational reasons, such as maintenance logs, is not easily summarized and analyzed in a way that is useful for decision making.  Rocky Mountaineer provided a student team with  five years of railcar maintenance records, much of it unstructured text, with thousands of records and around fifty variables, and no data dictionary. The team created a dictionary, then structured enough of the data to find useful patterns: for example the types of railcars and the amount and type of maintenance to be expected in the future.  They also recommended procedures and tools to further refine the existing data, and to improve data collection and quality going forward.  The work was useful enough that the student team was invited for a second presentation to a larger Rocky Mountaineer audience.

Software Bugs in Game Development

In the year-long process of developing a new game, Electronic Arts developers maintain records of bugs found in games.  EA gave the student team these records for four games and a very broad mandate:  look for correlations, trends, and data gaps; and make recommendations on EA's current status in collecting and using the information.  Organizing the teamwork in analytics project is always a challenge, and in this case especially so.  With four games, each team member worked on one, and had to regularly share results and insights with the other team members.  The result was an impressive, and coherent presentation to EA management.

Identifying Potential University Donors

Contacting alumni and other potential donors takes time and money for university fundraisers.  Using historical data of donations, and merging with external census and psychographic databases, the student group developed models of the probability of an individual with a specific demographic and psychographic profile becoming a major donor.  This allowed the fundraising group of the university to allocate their communication efforts according to the prioritized list of their potential donors.

eCommerce and Weather

The technical outdoor clothing and equipment manufacturer, Arc'teryx, believed that their online sales were likely related to weather, and wanted to know if it was, and how, to determine if they should watch weather forecasts to manage content and promotions.  They provided the student group with three years of transaction records from around the world, over four million records.  The student team had to locate historical weather records, decide on which locations to work with, what particular weather data might plausibly be relevant, and prep and clean both the transaction records and eCommerce records so that they could be merged and, ultimately, build over 450 predictive models.   The result:  no statistically significant relations were found.  A great example of a useful null result, and one that management found useful, so much so that the students were invited back for a second presentation to a larger audience at Arc'teryx. 

Hospital Fundraising

Two student projects involved analyzing several of the anonymized donor databases of the Vancouver General and UBC Hospital Foundation, and integrating with external census and geo- demo- and psychographic data.  Students developed profiles of different donor classes, developed predictive models of donation probability,  and on that bases provided recommendations on identifying and reaching potential donors. 

Emergency Room Admissions

Vancouver Coastal Health, a regional health authority, provides health care services in part of Greater Vancouver including Vancouver General Hospital.  Recently, VGH has experienced an increase in Emergency Room visits by patients with mental health and addiction complaints and wished to explore possible correlates with this increase.  One data set on anonymized emergency room visits was augmented with VCH Health Service client data.  In addition, students collected geodemographic data from Statistics Canada and geopsychographic Prizm data to explore spatial patterns. The data was cleaned, integrated and split at the time point where patient admissions had increased.  Descriptive visualizations and cross-tabs identified several useful correlates, which VCH operations managers found to be very insightful.

Graduate Employment

The Career Management Centre at the Beedie School of Business offers consultation to help Business students identify and achieve their career goals post-graduation. The Centre aims to maximize its benefit to students by optimizing its mix of program offerings. In order to determine the impact of the services offered by the CMC, four datasets were consolidated and analyzed to identify possible internal and externals factors that impacted student employment outcomes. By identifying specific attributes that lead to success in obtaining employment, the CMC can structure their programs to increase the number of students able to acquire a job after graduation.

Retail Product Mix Prediction

A provincial government liquor authority provided 40,000 transaction records sampled from all of the retail outlets in the province. Stores were clustered into six types based on product mix sales. Geodemographic and psychographic data were used as predictors in a multinomial model of this store type. The model can be used to predict the probable product mix demanded at any potential new retail location in the province based on the publicly available characteristics of the neighbourhood.