by Jared Norman
The 52nd Annual Conference of the Operations Research Society of South Africa was held at Pecan Manor Conference and Function Venue in Hartbeespoort. This year's event was hosted by UNISA. The venue maintains a forest of pecan trees and many animals casually grazed nearby as we gave our talks.
The conference began on Sunday afternoon with a delicious buffet - the first of many. It was lovely to see familiar faces from last year and catch up. With coffee to prevent the post-lunch slump we headed to the conference tutorial hosted by Elytica. My background does not include Mixed Integer Linear Programming but I appreciated the vision behind this software and enjoyed wrestling with the problem which would eventually be solved by Hendrik Snyman earning him a R3500 cash prize.
Monday morning began with a breakfast buffet (I told you there were many). Registration continued to take place and I was excited and terrified to see an interesting shaped Rubiks cube in my pack. Sharp also sponsored a calculator which does machine learning (linear regression). There were also some pens and a USB drive in there which are always useful and the bag itself was a handy container for pecan nuts. We were welcomed by the LOC Chair, Phillip Mashele, and addressed by ORSSA president David Clark. I particularly enjoyed the opening remarks of Prof. Marcia Mkansi which would set the stage for much of the conference and in particular a talk on the last day about the future of quantitative sciences in South Africa. The opening keynote by Prof. Ramasamy Sivasamy took us through a Jupyter notebook on how to detect heart murmurs (but also cuckoos) using Python.
This year, for many reasons, there was a single track running through the conference. This was refreshing because it meant nobody missed out on any talks and you always knew where you needed to be. The first session, Optimisation, saw four interesting talks. Funanani Netshitungulu talked about his multi-objective evolutionary algorithm approach to a generalisation of bin-packing whereby some items can be optional. Michaela Budge discussed her research into optimising the Formula 1 timetable such that distance travelled is reduced together with the very real world considerations one needs to take into account. Zachary Bowditch gave a talk on a very interesting real world problem whereby a particular type of hash is sought together with his results. The concluding talk by Jared Tavares was on solving CVRPTW using a hybrid memetic algorithm.
With optimisation out of the way it was time for another buffet. I enjoyed the discussions over lunch which took place outdoors.
The second session consisted of a single talk detailing Robert Bennetto's OR approach to solving his housing estate's loadshedding woes. The talk took us on his journey moving from problem to solution and I feel it really embodied the spirit of OR. We then had from 3pm to 7pm to enjoy the day. Personally I took the time to enjoy the grounds which included sampling of pecan nuts and a moment where Lieschen was chased by a zebra.
The evening greeted us with glasses of social lubricant and a roaring fire adjacent to the animals. Many conversations and catch-ups were had and again we were treated to a buffet of fruits, cheeses, and things on sticks. I reflected a bit on how things were during the height of COVID-19 and became grateful for this opportunity to network again.
Tuesday, after breakfast, we were treated to some of the newer OR methods in the topic of “Decision Support”. Emma Davis discussed her work analysing trending topics on Twitter using various measures of centrality to understand social influence. Jessica Rees gave us some interesting insights from her work at Discovery running Large Language Models in a POPIA compliant way. Jess would later be awarded the prestigious Category IV Recognition Award. Well done Jess!
Session four was for the National Student Competition in which honours and masters level students compete for sought after prizes. The first talk by Esranel Becker was given by Lieschen Venter who reported on an analysis of Sermon on the Mount using Game Theory. Marlize de Villiers talked about her decision support tool which was developed in addition to a model of optimal vehicle routing which considers driver-route familiarity. This would later result in the best Honours prize. The best Masters prize would go to her co-supervisor Jacobus King who remotely presented his work on solving the newly posed familiarity vehicle routing problem (FVRP). The last talk was also remote and given by Bianca van Zyl who looked at aspect-based sentiment analysis (ABSA).
The final session on Tuesday was on “OR in Education in Healthcare”. Here we saw Sha-abaan Slamang talk about using both a system dynamics model and agent based model for improving student graduation rates. We then had Hannah Jarvis talk to us about her work optimising mobile clinic schedules in rural areas. I was pleased to see the extent of involvement with the people who will actually benefit from this research in the research process itself. Finally I presented some of the research software engineering work I’ve been involved in at the Modelling and Simulation Hub, Africa (MASHA) which focuses on infectious disease modelling for policy.
One AGM later and we were at the Conference Banquet which took place in a pretty venue a short walk away. I enjoyed the speeches which were given and enjoyed celebrating the success of the award winners, many of whom I have already mentioned. I was quite involved in discussion at the event but I do recall seeing some dancing in my peripheral vision.
The final day started with session 6 which was for “Miscellaneous” talks and began with Modisane Seitshiro discussing his work in ARIMA misspecification analysis. The topic brings with it the question of what a model is and how to use one responsibly which is arguably applicable to everyone in OR. Kenneth Dukuza gave a talk on the use of a nonstandard denominator in deriving an exact solution to stochastic differential equations. Lieschen Venter, who runs the Systems Thinking for Education Policy (STEP), gave a talk on using Game Theory to optimise for romantic happiness. I hope to see future work in this area modelling agents moving around Die Mystic Boer in Stellenbosch. We concluded with a talk from Robert Benneto which linked to the problem Zachary Bowditch reported on. This problem was about clustering districts (polygons) to service waste-truck disposal using multi-objective criteria.
We had two final talks after tea that Wednesday. The first talk was given by former VC at Wits Loyiso Nongxa and president of SASA Inger Rotelli. They described their role in an initiative looking at improving the outlook for quantitative sciences in South Africa. The final talk of the day was from Anna Bershteyn from IDM who discussed mathematical modelling in health. She gave a high-level overview of the field and surveyed her experience working on optimising for equity as well as efficiency.
The best masters medal is known as the Theodor Stewart medal. Theo was unable to attend this year but he turned 80 on the Saturday proceeding the conference so we recorded a happy birthday message for him.
After the concluding remarks I was happy to enjoy one final lunch buffet before heading to the airport. I look forward to seeing everyone at the next one, if not before that.
The gallery of photographs for ORSSA 2023 is available here.