D3EMO Modeling Week – Data Driven Decision Making and Optimisation

The D3EMO Modeling Week on Data Driven Decision Making and Optimization took place on 3rd-7th September 2018 at the Copenhagen Business School (CBS) in Copenhagen, Denmark

It was organised by

–       Prof. Dolores Romero Morales, Professor of Operations Research, Copenhagen Business School, Department of Economics

–       Prof. Emilio J. Carrizosa Priego, Professor of Statistics and Operations Research, University of Seville, Department of Statistics and Operations Research

Problems were presented by

  • AirFrance (Augustine Lombard)
  • Central Bureau voor de Statistiek (Marc Ponsen)
  • DSB (Natalia Rezanova)
  • TESCO (Eleanor Evans and Matthew Fryett)

PhD, PostDoc and researchers from all over the world (Belgium, Denmark, Greece, Italy, Poland, Spain, UK, Vietnam, etc.) attended this modelling week.

On Monday 3rd, after the welcome given by organizers, two companies (TESCO and CBS) presented their problems and the seventeen participants were divided into two groups, one for each problem; we had time until Tuesday evening to study the problem and propose some solution methods. On Wednesday 5th the other two companies (DSB and AirFrance) introduced their two problems and two new groups were formed, which worked together until Thursday evening. 

Companies’ problems can be classified into two macro-category: dealing with a lot of data (forecasting and visualization) and reasoning about the abstract problem (modeling).

During Friday, the last day of the Modeling Week, every group exposed its results, through an oral presentation and some slides; DSB representatives were present. Moreover, also some trainees talked briefly about their research topics. The study group ended in the early afternoon.

All groups analysed deeply the companies’ problems and produced written reports or slides, which will be delivered to the companies. In particular:

–       The TESCO group tried to replicate the company model to forecast the customer daily demand for each product and store of the firm; they implemented some forecasting models after analysing the dataset through queries and statistical methods;

–       The CBS group studied the dataset about air pollution and proposed methods to visualize data in order to have more sociological impact;

–       The DSB group modelled the problem of scheduling cleaning operations of trains in the same way of the company (except for some simplifications), they developed their own instances and also suggested other models and ideas to tackle the problem;

–       The AirFrance group tried to model the planning of aircrafts but had to deal with constraints very difficult to describe mathematically.

Another main outcome the whole modelling week produced was making all of us trainees bond with each other, creating connections that will last in the years to come, as well in other conferences or PhD schools. It was also great and inspiring to see companies really interested to collaborate with universities and research centres, giving us the chance to work on real problems.

By Alice Raffaele


About the Author:

Alice Raffaele is a PhD student in Operations Research from Italy. She is finishing her first year and is interested in OR at 360 degrees. From modelling to polyhedral analysis, but mostly its applications to industrial problems and the interplay between optimisation and data science. She enjoys educational mathematics, i.e. activities and games for children attending primary and secondary schools, to make them discover the beauty of maths and problem solving.


Second Agri-Food Study Group with Industry

The Second Agri-Food Study Group with Industry will be held at ICMS in Edinburgh on the 21st – 23rd February 2018.
If you wish to take part, please register here.
Clean Growth, and the AI & the Data Economy are two of the four Industry Strategy Grand Challenges underpinning much of HMGs investment into R&D over the next few years. Additionally, we can expect  funding in Transforming Food Production to be made available as an Industrial Strategy Challenge Fund, this Study Group is a key way of establishing relationships between communities, refining industry challenges and preparing the mathematical science community for this upcoming opportunity.
The three problems being presented are:
Promar International – Identifying Drivers for Profitability in Cattle
We have a huge amount of data throughout the dairy farm supply chain. We have used this data predominantly for benchmarking, including the impact of different farming systems and geographical areas on profitability. We have done some analysis to identify drivers for profitability using physical and financial parameters but more recently management practices and attitudinal aspects of the farmers. Based on the datasets above, we would like to explore drivers and KPI’s to predict profitability (performance is often masked by the management ability of the farmer and other factors). Another potential area for exploration is in linking genetic and financial data on an individual cow basis. 
Phytoponics – Aeration Optimisation
Phytoponics Hydrosac is a hydroponic growing system module that holds a body of water to grow plants in. At the base of the module is an integrated aerator, which consists of a perforated strip of material that receives external air input from an air compressor, and emits bubbles to the body of water such that oxygenation of the water occurs. The scope of this challenge is to develop a mathematical model of the aeration system of the Hydrosac, including volumetric flow rate, input pressures, aerator strip material design parameters and costs therein, such that Phytoponics can use this model to improve the aeration of the Hydrosac design and select supporting ancillary air supply services or system parameters.
Syngenta – Scheduling Seed Production
Syngenta are one of the largest suppliers of agricultural seed globally. A key requirement of the business is the adequate supply of seeds to meet varied customer demands throughout the world. Scheduling seed production is complex and unpredictable. Crops must be planted one year in advance of when the resultant crop of seeds will be sold. A recurring problem is that of spatiotemporal variation of yield and the management of the associated risk of over / under production of seeds, which is extremely costly, and can severely damage the business. Syngenta have developed an interface for internal planning of production, which is purely based on historical yield. Syngenta would like to rationalise planting strategies which are informed by a judicious choice of objective function, which best optimises the business performance (which could include growth, profitability) and is robust against potential risks (natural, market risks etc). Can a more sophisticated approach “beat” the experts and / or strategies based on historical data simulations?

More information on these problems can be found on the Study Group website. 

The KTN staff and members of their Industrial Mathematics group worked with the University of Bath’s Institute for Mathematical Innovation (IMI) to organise a study group. The Agri-Food Study Group brought together over 40 mathematicians, engineers and computer scientists to work on challenges presented by representatives from the Agri-Food sector over the course of three days. The Study Group which ran from 16-18 January 2017, and was hosted by the IMI and sponsored by Innovate UK.

Three agri-food challenges were presented at the event, namely helping farmers to optimise the value of the pigs they sell (Innovent Technology) improving cocoa yields for the chocolate industry (Mondelez International), and refining the design of a hydroponics system for crop production (Phytoponics Ltd). Further details about these challenges are detailed here.

These challenges required varied expertise from across the mathematical sciences, and it was fascinating to see the three agri-food company representatives working closely with the maths experts over three full days to try and solve the problems presented.

On behalf of Dr. Matt Butchers, Knowledge Transfer Manager, Industrial Mathematics 
Knowledge Transfer Network