Improving teacher insight
With more than 100.000 students in primary school using Clio Online’s platform every day, their log files contain insightful data about students’ behavior, results and learning progression. Efficient algorithms and analytical tools provide the teachers and schools with this information based on the lessons assigned to the students. These data are valuable as they can improve teacher insight for the benefit of the individual students.
Predicting a student's score in a quiz/exercise is one of the important tools that we are currently developing. A reliable prediction can help teachers identify weak students that require special support, generate adaptive hints, and improve students' learning. More precisely, we develop an algorithm that solves the following problem: Given the results on quizzes the student has already completed, it can predict how well the student will answer another quiz.
We formalize our prediction problem as the matrix completion problem and investigate variants of weighted low-rank matrix approximation (LRMA) for solving it. Currently we are investigating several efficient iterative methods with guarantees for the weighted LRMA. Furthermore, we make experiments to see what kind of log data are most important for predicting a student’s performance.
We expect that the project will result in knowledge about efficient algorithms for improving teacher insight. Furthermore, by identifying what log-data are important for predicting a student’s performance, we will gain knowledge about the connection between student behaviour and the rate of learning.
The solution will offer a leap in state-of-the-art support of teacher insight and addresses key customer demands. Combined with adaptive algorithms, the solution can be used to tailor the learning material to the student’s optimal learning pattern. At the same time it can improve teachers’ knowledge and job satisfaction, create more motivated students, and increase the quality and attractiveness of the primary schools.
This project is a joint project between Department of Computer Science, University of Copenhagen and Clio Online.