Saturday, April 16, 2016

Fabulous Analytics Day at the University of Waterloo in Canada

This past Thursday, I flew (after a two hour delay) on Air Canada from Bradley Airport to Toronto since I was to speak at Analytics Day at the University of Waterloo. The conference was organized by Dr. Fatma Gzara, who is a faculty member in the Department of Management Sciences there and who had also hosted me almost exactly 5 years ago for a seminar!

I accepted the invitation to be a plenary speaker at this conference and the experience was tremendous plus I always enjoy not the flight delays but the serendipity of whom you may meet while traveling.

I was put up in the stunning Delta hotel, which opened about two yeas ago,  and not only was I staying at the hotel but Dr. John McCarthy, who is the Graduate Dean at UMass Amherst and a Vice Provost, was also there. I found this out while talking to two academics in the elevator and went to see him at a breakfast. I had seen John only two days prior at the UMass Honors banquet. He was in Waterloo for the annual meeting of the Northeastern Association of Graduate Schools!

And there was another luminary staying at the same hotel, whom I had seen Thursday evening on the Canadian news - none other than Prime Minister Justin Trudeau. I had thought that all the nicely dressed men were actually hotel staff but they were security guards. I saw Trudeau not only yesterday morning as I was getting picked up to go to the conference but also when I returned to the hotel. I took the photo of him entering the limo below. Since I was born in Canada, this was extra special.
My plenary talk was on: Predictive and Prescriptive Models of Cybercrime an Cybersecurity Investments Under Network Vulnerability. The conference was well-advertised and attracted practitioners as well as academics from numerous universities in the Toronto area. Dr. Gzara runs the Waterloo Analytics and Optimization Lab. She had taken part in the recent INFORMS Analytics conference in Orlando and in her presentation included the following slide which was generated by a text mining program of all the abstracts at the INFORMS conference.
Great to see Optimization Models prominent  in the text cloud along with Big Data and Supply Chains!

There were opening remarks by the Associate Dean of Engineering at Waterloo. The full conference schedule can be viewed here.

Dr. Gzara did a wonderful job of introducing each speaker and had a slide prepared on each.

I very much enjoyed the talk Optimizing Supply Chains with Advancd Analytics presented by Pete Devenyi and Scott Wahl of Dematic North America.

The Detamic presentation was on electronic commerce and grocery distribution centers. Dematic is a global software engineering firm with 6,500 employees and $70 million spent on R&D a year (I assume that is in Canadian dollars). Their clients include Tim Hortons, Amazon, and Walmart, so I enjoyed speaking with them about my experiences last October at Amazon's first Supply Chain Optimization Conference in Seattle. They talked about predictive analytics to get items to customers in 1/2 a day and placing warehouses closer to customers. I very much enjoyed their presentation which had videos and was very dynamic. They noted that "warehouses" in the future will be truck.

Since there were both industry and academic talks the discussions were great and the day flew by much too quickly. The Dematic presenters stated that they are trying to solve the problem "of the tyranny of now". Customers when ordering online want their orders to be delivered asap. Without the right analytics firms don't have visibility, which is critical in e-commerce.  Dematic also uses descriptive analytics to figure out bottlenecks and is looking towards work force optimization and item storage optimization. Visualization analytics can reveal where there are bottlenecks in terms of processes in a grocery distribution center, for example. Grocery distribution centers are completely different from e-commerce ones, since grocery orders are placed every day. It may cost $100 million to automate a grocery DC. There is even pallet optimization that needs to be done.

The third presentation was by Chad Elkadri, Reza Farahani, and Hany Kashif of SOTI.
As you can see from the photo below, the audience was very attentive and interested in the talks.
SOTI develops software for enterprises and was founded in Canada 20 years ago. It, through its Research and Innovation Lab, has been working on the Internet of Things. Specifically, the company focuses on managing devices including connected devices in the enterprise. I was also very interested in this presentation because only 10 days ago, my collaborator, Professor Tilman Wolf, presented our latest work on the Internet of Things at the IEEE Berlin conference.  

SOTI builds partnerships with OEMs and is in 170 countries. Partners include device manufacturers and also operating system companies such as Google, Apple, and Microsoft.
SOTI speakers noted how sensors are paving the way for Big Data and it is essential to recognize the varying levels of the importance of data. Also, there is "demand for new analytical models." SOTI is building a performance-aware engine, which was the topic of the presentation. Also, one of the noted challenges is how to extract data without changing the behavior of the system. They also presented a model, which I liked, which had a parameter that captured the value of a data element at a point in time, and I asked a question about this. Performance-aware analytics must take constraints into consideration. Such constraints can include QoS (Quality of Service) constraints. The case study that was presented was on connected cars.

Then it was time for lunch which included all sorts of wrap sandwiches - I chose the salmon one as well as salads and fresh fruits plus a tray of sweets.
After lunch, an academic talk was given by Joe Naoum Sawaya of the Ivey School at the University of Western Ontario. His title was: Optimization in the Age of Modern Analytics and Internet-of-Things. Saway discussed: descriptive, predictive, and prescriptive analytics and emphasized that now algorithms need to be intuitive, distributed, scalable, and have a niche or application. He talked about sensors for Smart Cities since he had worked for IBM for 5 years.  He focused on the automotive industry.  He presented a nice model focusing on SPONGE, which is an energy matching supply and demand framework but that ensures privacy since the central authority does not now the individuals' utility functions.  He noted that because of the Internet of Things, we no longer can assume centralized decision-making.

The second presentation after the lunch was by the Deloitte Consulting Canada group. The speakers were: Snehar Patel and Niraj Dalmia and a consultant. Since I have had one of my former students work for Deloitte (Christina Calvaneso) I also was looking forward to this talk.
They spoke on use cases including one in the oil and gas sector, in which there has been tremendous economic upheaval. Hence, figuring out where costs can be reduced is imperative. They talked about system dynamics approaches (top down) and also agent-based modeling (bottom up).  They noted that "analytics is an opportunity." One example was a mine in Chile in which they use predictive analytics to prepare for possible disruptions.

Another example, which they spoke on, which reminded of a project that some of my Opertaions Management students did a while back in my Transportation and Logistics class at the Isenberg School  and that was on modeling patient flow from an emergency room in a hospital.  Deloitte is taking data from clients and then building models and this is the direction that consulting is going in they told the audience.

Dr. Gzara, the organizer, was the last speaker of the day and gave an excellent talk on working with one of her industry partners - Dematic. She spoke on her group's work on using data from a client's warehouse (and the data is by no means complete) to do diagnostics and to evaluate bottlenecks and where the system processes might be improved. She presented a series of innovative models and also spoke on how descriptive analytics showed them that there is a high variability in orders and some orders have as many as 31 products in them.She discussed network flow models and also a model that I very much liked - using community detection to also graphical depict orders as a network. How can one use community detection and network learning to identify which products are co-ordered. She also stated the need to do predictive dynamic analytics for product order groupings over time.

I made sure that a photo was taken of me with Fatma for a keepsake of an intellectually rich and very stimulating day.
My trip via a shuttle to the Toronto airport was pleasant and there was not so much traffic although it was rush hour.  And, would you believe, I took Route 401 both to and from the Toronto airport to the Delta hotel in Waterloo and the shuttle driver then informed me that the route, Route 401, is the busiest freeway in North America!  I also saw signs of $10,000 penalties if a driver exceeds the speed limit by 50 kilometers (which is 30 miles) and the car also gets confiscated by the police for 7 days. This is to prevent drag racing but, given the congestion, drag racing would be a challenge in itself.

And, as one would expect, I had a delay on Air Canada on my flight back to Hartford. So I took a photo at the airport of one of the national foods of Canada and that is the Tim Hortons donuts.
For the flight back I purchased the last copy of The New York Times and in it was an OpEd piece co-authored by my Provost at UMass Amherst, Dr. Katherine Newman!  My day began with a UMass administrator and in a sense ended with one. The academic network is global and far-reaching and wonderful, I must day.

Many thanks to Dr. Fatma Gzara for a fabulous Analytics Day!