With the successful application of machine learning in various domains, it is not surprising to see a growing interest in the integration of learning capabilities into solving combinatorial optimization problems. The most considerable interest is in routing problems (mainly for the TSP and VRP), as they have been widely studied by the combinatorial optimization community. However, the application of machine learning models is not limited to being part of the solution process but can also be used to identify implicit knowledge of real-life route deliveries. For example, it has been found that skilled drivers do not necessarily follow the routes suggested by route optimization software. Instead, they consider other practical aspects that they have accumulated through experience, allowing them to take routes based on road conditions, traffic patterns, parking convenience, and other considerations.

By being able to detect patterns in the routes actually used, machine learning tools can be useful to support optimization methods that aim to find high-quality practical solutions to routing problems. To be able to detect tacit knowledge in routing problems, a first step that will be covered in this thesis is to develop a data-driven approach that can identify the formal constraints of a routing solution. Given a set of routes, can a machine learning model be used to accurately detect constraints that are hidden on purpose, such as hard constraints regarding time windows, precedence, inventories, or capacities?

The following main tasks could be part of the thesis work:

•  Study the relevant literature that uses machine learning to solve routing problems

•  Create a training set of routing solutions for different variants of routing problems

•  Develop and implement a machine learning model

•  Evaluate and analyze the ability of the model to identify formal constraints

This topic requires excellent skills in programming and the use of machine learning tools.

Collaborators: Researcher Mohamed Kais Msakni, Professor Peter Schütz (both NTNU)

This research problem is given by the Norwegian Labour and Welfare Administration (NAV), which is an organization that administers a third of the national budget through schemes such as unemployment benefit, work assessment allowance, sickness benefit, pensions, child benefit and cash-for-care benefit. 

The problem is presented as received:

"Hvordan offentlige oppdragsgivere kan få til innovasjon i kontraktsperioden?"

The length of stay (LOS) is an important indicator of the efficiency of hospital management. Reduction in the number of inpatient days results in decreased risk of infection and medication side effects, improvement in the quality of treatment, and increased hospital profit with more efficient bed management. The purpose of this thesis is to determine which factors are associated with length of hospital stay, in the covid period using the collected data. In this thesis the student will apply a data mining approach to understand the management process in the covid period. The data is from a Portuguese hospital.

Process mining is a data analytics approach which has shown promising results in healthcare including the potential to improve the management of chronic diseases such as cardiovascular disease (CVD). Cardiology is a branch of medicine that diagnoses and treats heart and blood vessel illnesses include coronary heart disease, cerebrovascular disease, rheumatic heart disease, and other heart conditions. CVDs are the world's leading cause of disease-related death. In 2015, the top two cardiovascular diseases, coronary heart disease and stroke, caused 15 million deaths worldwide. From an economic standpoint, CVD has a significant impact on healthcare expenses, productivity loss, and the care of persons with chronic illnesses. A big problem is lowering the cost of CVD care while also enhancing the quality of care.

Helping healthcare professionals develop a better understanding of how to improve CVD care pathways may result in better outcomes for patients.   A Data analytics process will be applied to the collected big data from a hospital. Process mining can be applied in healthcare settings to give new insights that help enhance patient treatment processes (also known as care pathways).

FRAM is the brand name of Møre and Romsdal (M&R) counties’ integrated public transport service system. The system combines multiple ferry-, express boat-, and bus services, with the latter accounting for some 9 million passengers and ticket revenues of 170 million NOK per year.

FRAM offers an extensive set of ticket options, so travellers can pick the best fare that fits their individual needs. Among other things, passengers can choose between paying onboard (e.g., cash/credit card) or using off-bord payment solutions (e.g., FRAM-app) to purchase tickets before the trip. From an operational perspective, FRAM prefers off-board payments as they reduce the bus drivers’ workload and increase the average system speed. FRAM currently charges a 20-kroner service fee for all onboard tickets to incentivise travellers to use off-board payment options. Despite this premium, some 40 per cent of our customers still purchase their tickets on board, suggesting that some traveller segments might value the option to pay onboard higher than 20 kroner.

A successful MSc-project analyses passengers’ valuation of onboard/off-board payment options in M&R and proposes adequate surcharge levels for different traveller segments.

The analysis is based on applying a suitable (preferably quantitative) research methodology (e.g., see LOG904-148/LOG904-150) and is closely coordinated with FRAM. We invite interested students with an average grade of ‘B’ (or better) and (at least) intermediate language skills in Norwegian to get in touch with edoardo.marcucci@himolde.no and/or falko.muller@mrfylke.no.

Star Information system is a 25-year-old Software Company with main office in Trondheim. Star deliver procurement, logistic and maintenance software primarily to maritime and offshore sector. Star have customers all over the world among others Color line, Van Oord, COSL, OKEA and Altera. One of the main fields Star Information System is working in is spare parts and inventory, centralized/decentralized inventory, SCM etc. for the customer of Star Information System. Star Information System will provide data.

This master thesis may be approached by using optimization methods or general SCM, which means that it could be relevant for students from both directions.

SIRQEL is a Norwegian based startup, which provides rental service for families with young children for example, bicycles. Customers can order products from our website and get it delivered at their doorstep. The service is subscription based and customers can change the size if it doesn’t fit or send it back to us for repair if something gets torn or broken. We are responsible for any repair and washing before the product collected is sent to a new family. Thus, it can be said that this is a Product-As-A-Service supply chain. Our aim is to replace single use purchase and make it easier for families to adopt sustainable consumption. However, through different tests and trials, some products are rented out more than others.

We want students who are interested in Circular Economy and Service logistics to help us with conducting research into one or more of the following topics:

- Identify the barriers and ways to mitigate them while transitioning traditional product categories into a product-service model.

- Explore which other product categories can be turned into a service model.

- Investigate if PAAS for bikes is better in regards to CO2 emissions.

- Currently, the repair and delivery are outsourced to another service provider. Explore whether one or both services can be managed in-house, in terms of costs and lead time.

- At present, we rent out the products we have already bought and is in our storage. Thus, it is limited and there’s a chance we might be understock if all products are rented out and we get a new order. However, buying a lot of inventory and storing them will also have some costs. Explore whether Just-in-Time inventory can be implemented, its challenges and ways to mitigate them.

Apart from these, students are also encouraged to propose their own research idea which might help the company.