Bayesian Knowledge Tracing For Assessment Results Analysis
This was a collaboration research project between Uniparticle, where I worked as an ML research
engineer, and Discovery Education.
In this paper, A Bayesian Knowledge Tracing model ( a Hidden Markov Model) is utilised to
perform analysis on assessment results to get human-free feedback on the quality of the
questions.
The model is built and trained using PyBKT on the National Coding Competition 2018 records,
obtained with the help of Discovery Education.
The technique uses guess and slip parameters of the model to perform the analysis process. The
details of the recorded data analysis, the assumptions made to build the model, the training and
the validation, and the obtained results are all discussed in the paper.
The method was proven to help detect some problems in the given questions.
This paper was part of my work at Uniparticle. The paper was reviewed by the
committee of NCC, discussed on 2021 17th International Computer Engineering Conference (ICENCO),
and published on IEEE.