Computer Science

Faculty of Engineering, LTH


CS MSc Thesis Presentation Day August 25 2022


From: 2022-08-25 09:15 to 18:00
Place: See information for each presentation
Contact: birger [dot] swahn [at] cs [dot] lth [dot] se
Save event to your calendar

Four MSc theses to be presented on Thursday August 25, 2022

Thursday August 25 is a day for coordinated master thesis presentations in Computer Science at Lund University, Faculty of Engineering. Thirteen MSc theses will be presented.

You will find information about how to follow along under each presentation. There will be presentations in two different rooms: E:2405 (Glasburen) and E:4130 (Lucas). A preliminary schedule follows.

Note to potential opponents: Register as an opponent to the presentation of your choice by sending an email to the examiner for that presentation ( Do not forget to specify the presentation you register for! Note that the number of opponents may be limited (often to two), so you might be forced to choose another presentation if you register too late. Registrations are individual, just as the oppositions are! More instructions are found on this page.


E:2405 (Glasburen)



Presenters: Lykke Axlin, Klara Broman​​​​
Title: Identification of relevant error descriptions in build logs using machine learning
Examiner: Elin Anna Topp
Supervisors: Martin Höst (LTH), Marcus Klang (LTH), Gustaf Lundh (Axis Communications AB), Ola Söder (Axis Communications AB)

The Jenkins plugin Build Failure Analyzer (BFA) is used to help engineers find lines containing relevant error descriptions in logs from failed builds. The BFA utilises manually crafted regular expressions to scan the build logs. As new error descriptions arise in the build pipeline, new handcrafted regular expressions need to be added, which is a time-consuming process. This thesis aims to investigate to what extent it is possible to use historical data from the BFA to train an AI-model to detect relevant error descriptions that are not currently found by the BFA. Further, we intend to examine what requirements such a model needs to fulfill in order to be useful at the case company. The best performing model achieved an F1-score of 0.60. Further improvements to the dataset or a more advanced machine learning model is suggested before the model can be used as a complement to the BFA.

Link to popular science summary:

11:15-12:00 (also via Zoom)

Presenter: Martin Hansson Tomicic
Title: Return on investment and maintenance in a Mobile test automation implementation
Examiner: Per Runeson
Supervisors: Masoumeh Taromirad (LTH), Niklas Kenéz (Robert Bosch AB)

Bosch SensorTec develop MEMS-sensors and corresponding software for many uses. The thesis consists of researching and implementing a test automation solution for a mobile application that interacts with a sensor. Research is focused on the current state of testing, the team's concerns regarding the implementation, as well as regarding return on investment and the maintenance of the proposed solution. Results include a list of requirements for the solution itself, as well as a working implementation that is connected via Jenkins to version control. Maintenance is presented via measurements of executions on several versions of the application as well as how the solution will handle different types of test cases available. Return on investment is defined for the particular project and is shown to be feasible both through improvements in execution time and possibility of integration in a pipeline as well as enabling the testing team to focus on system testing and other tasks.

Link to Zoom presentation:

Link to popular science summary:


E:4130 (Lucas)



Presenters: Viktor Karlsson, Aston Åkerman
Title: Real-time detection of spelling mistakes in handwritten notes​​​​​​​
Examiner: Marcus Klang
Supervisor: Pierre Nugues (LTH)

The art of handwriting has been relevant for thousands of years and continues to be so in modern times. Technology that can recognize handwritten text in images could be utilized by smart-glasses, as a supportive tool for visually impaired or as a tool to facilitate learning how to write. With this thesis, we present "Ortographer", an end-to-end handwritten text recognition system, able to detect spelling mistakes in handwritten text in real time. The system shows promising capabilities, detecting ~69% of the spelling mistakes in our test set and running at ~17 ms per word.

Link to popular science summary: TBU


Presenters: Asmail Abdulkarim, Filip Johansson​​​​​​​
Title: ML-driven self-tuning MySQL-databases​​​​​​​
Examiner: Pierre Nugues​​​​​​​
Supervisor: Luigi Nardi (LTH)

As databases has become more critical part of our digital infrastructure, the the cost of maintaining a database has increased. Database management systems such as MySQL allow users to tune parameters or knobs as they are called in order to improve performance. This is a difficult optimization problem, especially for humans to do manually. This is because there are hundreds of knobs with complicated dependencies between many of them and because the optimal configuration depends on the application and also changes over time as the workload changes. For these reasons automating this process is of interest to many companies and database administrators. In this project we use an existing database optimization tool and investigate the possibility of defining a search space for MySQL capable of producing good configurations in reasonable time for any workload.

Link to popular science summary:




How to get to computer science