Skip to content
A Research Project on Bayesian Knowledge Tracing
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

All of Knowledge Tracing

Peter Swire -

What is it

A research project on bayesian knowledge tracing. You can see the paper in progress here.


Bayesian Knowledge Tracing is a simple, interpretable method for inferring a student's knowledge of a subject. Given a student's responses to questions regarding a single skill, it captures four probabilities: the probability the student knew the skill before starting, the probability the student will learn the skill, the probability they will guess a question right, and the probability they will slip and answer a question wrong. These are only four parameters per skill, but this is a lot: a tutoring system like ASSISTments can have hundreds of skills, and sparsity issues can lead to inaccurately-trained models.

This paper proposes a method to side-step the issue of training individual knowledge tracing models. The interprebility of Knowledge Tracing allowed us to overlay semantic constraints on the search space. These constraints let us then roll the entire search space into a single model, effectively making Knowledge Tracing parameterless.

How to use it

java -jar jar/main.jar


java -jar jar/predict.jar experiment_name train.csv test.csv

What it uses

JUnit - Common Public License

XStream - BSD

sqljet - GPL

You can’t perform that action at this time.