## Algorithms and AI in Education

Publikation: Bog/antologi/afhandling/rapport › Ph.d.-afhandling › Forskning

#### Standard

**Algorithms and AI in Education.** / Hjuler, Niklas Oskar Daniel.

Publikation: Bog/antologi/afhandling/rapport › Ph.d.-afhandling › Forskning

#### Harvard

*Algorithms and AI in Education*. Department of Computer Science, Faculty of Science, University of Copenhagen.

#### APA

*Algorithms and AI in Education*. Department of Computer Science, Faculty of Science, University of Copenhagen.

#### Vancouver

#### Author

#### Bibtex

}

#### RIS

TY - BOOK

T1 - Algorithms and AI in Education

AU - Hjuler, Niklas Oskar Daniel

PY - 2019

Y1 - 2019

N2 - In this thesis we show several new results in a broad spectrum, ranging from theoretical computer science to data analysis.A brief summary of the results are:One-Way Trail Orientations: We show that there exists a strong orientation of a graph if and only if the graph is two edge connected, even if the edges are partitioned into trails, thus improving Robbins theorem from 1939. Fully Dynamic Consistent Facility Location: We show how to maintain a constant approximation for the facility location problem for general metrics with running time O(log(n)n) and total recourse O(n). Detecting ghostwriters in high schools: Author verification, i.e. the task of verifying if an acclaimed author of an essay actually is the author. This is a known problem in the Natural Language Processing community. What is new, is doing this in the educational setting, challenging whether the student has actually written the assignment in question. With our method on a balanced data set, we achieve an accuracy of 87.5 percent. Investigating Writing Style Development in High School: For every student we make a profile by comparing how similar a current assignment is to previous assignments. The profiles are then clustered and analyzed. Furthermore, we compare how the average similarity evolves over time. Sequence Modelling For Analysing Student Interaction with Educational Systems: Using log data we model student behavior as a distribution of Markov chains. The Markov chains are analyzed, and 125.000 of the sessions are deemed suboptimal from a learning perspective. Tracking Behavioral Patterns among Students in an Online Educational System: We make a soft clustering of a students activity during one week, using log data from an educational system. Based on the results, we give suggestions for an improved learning experience for the students. DABAI: A data driven project for e Learning in Denmark: In this paper we give an overview of the goals of the projects involved in DABAI, and what issues the educational companies are interested in.

AB - In this thesis we show several new results in a broad spectrum, ranging from theoretical computer science to data analysis.A brief summary of the results are:One-Way Trail Orientations: We show that there exists a strong orientation of a graph if and only if the graph is two edge connected, even if the edges are partitioned into trails, thus improving Robbins theorem from 1939. Fully Dynamic Consistent Facility Location: We show how to maintain a constant approximation for the facility location problem for general metrics with running time O(log(n)n) and total recourse O(n). Detecting ghostwriters in high schools: Author verification, i.e. the task of verifying if an acclaimed author of an essay actually is the author. This is a known problem in the Natural Language Processing community. What is new, is doing this in the educational setting, challenging whether the student has actually written the assignment in question. With our method on a balanced data set, we achieve an accuracy of 87.5 percent. Investigating Writing Style Development in High School: For every student we make a profile by comparing how similar a current assignment is to previous assignments. The profiles are then clustered and analyzed. Furthermore, we compare how the average similarity evolves over time. Sequence Modelling For Analysing Student Interaction with Educational Systems: Using log data we model student behavior as a distribution of Markov chains. The Markov chains are analyzed, and 125.000 of the sessions are deemed suboptimal from a learning perspective. Tracking Behavioral Patterns among Students in an Online Educational System: We make a soft clustering of a students activity during one week, using log data from an educational system. Based on the results, we give suggestions for an improved learning experience for the students. DABAI: A data driven project for e Learning in Denmark: In this paper we give an overview of the goals of the projects involved in DABAI, and what issues the educational companies are interested in.

UR - https://soeg.kb.dk/permalink/45KBDK_KGL/1ed7rpq/alma99123207868705763

M3 - Ph.D. thesis

BT - Algorithms and AI in Education

PB - Department of Computer Science, Faculty of Science, University of Copenhagen

ER -

ID: 231411931