PathAnalysis

This is the official implementations of On adaptive learning paths using hidden Markov models(JKDAS, 2023.02). For technical details, please refer to:

On adaptive learning paths using hidden Markov models [DOI]
Hyunhee Choi, Yunji Lee, Hayun Lee

(1) Setup

This code has been tested with R 4.2.2, Python 3.8, hmmlearn 0.2.8, CUDA 11.2 and cuDNN 8.0 on Ubuntu 20.04.

• Clone the repository

git clone https://github.com/LAIVDATA-EdTech-R-D/LearningPath-using-HMM.git

• Setup dependencies

conda create -n PathAnalysis python=3.8
conda activcate PathAnalysis
deb https://cloud.r-project.org/bin/linux/ubuntu focal-cran40/
sudo apt-get update
sudo apt-get install r-base
cd PathAnalysis

(2) AI-Hub dataset

AI-Hub(Mathematics learner's ability level measurement data) mathmatics dataset can be found here. We extracting the learner's log related with Addition & Subtraction. Data Format is:

Learner no. \ KC	KC01	KC02	...	KC07
1	1	0	...	1
2	1	0	...	1
...	...	...	...	...
183	1	1	...	1

(0: Incorrect, 1: Correct)

(3) Analysis

• KC Relation Selection using LASSO & RF

Rscript Elasso_RF.R

• Hidden Markov Model
  • HMM with KC Relation Selection

  python FeatureSelectionHMM.py
  

  • HMM without KC Relation Selection

  python OnlyHMM.py
  

• Structural Equation Modeling
  • SEM

  Rscript PLSPM_SEM.R
  

  • SEM for every possible paths

  Rscript PLSPM_SEM_Every.R
  

(4) Result(Graph)

KC prerequisite graph(KC Map)	Learning Path