Robotics and Semantic Systems

Computer Science | LTH | Lund University

Denna sida på svenska This page in English


CS MSc Thesis Presentation 18 August 2023


From: 2023-08-18 10:15 to 11:00
Place: E:4130 (Lucas)
Contact: birger [dot] swahn [at] cs [dot] lth [dot] se
Save event to your calendar

One Computer Science MSc thesis to be presented on 18 August

Friday, 18 August there will be a master thesis presentation in Computer Science at Lund University, Faculty of Engineering.

The presentation will take place in room E:4130 (Lucas).

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.)

10:15-11:00 in E:4130 (Lucas)

Presenter: Dan Svenonius
Title: Anomaly Detection with Machine Learning in OpenStreetMap Changesets
Examiner: Jacek Malec
Supervisors: Patrik Edén (LU), Hampus Londögård (AFRY AB)

OpenStreetMap is an open data set in which anyone can make contributions. This makes it prone to user errors in edits, and validation of these edits is re- quired to make the data more trustworthy. As there are vast amounts of possible errors, it is difficult to create a number of heuristics for validation, making a ma- chine learning approach interesting which could act as an initial filter, flagging potentially erroneous edits for further inspection. This thesis investigates the potential of machine learning in validating Open- StreetMap changesets. It builds upon existing work in vandalism detection but attempts to generalize the problem from vandalism to unintentional errors. It compares neural networks to tree-based models and finds that vandalism detec- tion methods with machine learning is effective in this task, where tree mod- els perform significantly better than neural networks. In particular, gradient boosted trees performs best with regards to every metric, achieving 89.46% accuracy. After surprisingly good performance, an extensive feature importance analysis is performed.

Link to popular science summary: