Comparative Experiment

---

Introduction

We use 8 models to compare with CAKT, namely DKT, DKVMN, SKVMN, SAKT, EKT, CKT, DKT-F (DKT+forgetting) and AKT.

Following the instructions, and you can get the results of these models quickly.

Environment

These projects are developed using

• python 3.7

• Pytorch 1.4.0

• Tensorflow 1.13.1

• visdom 0.1.8.9

• torchnet 0.0.4

• pandas 1.1.4

• tqdm 4.51.0

• numpy 1.19.2

• Pillow 8.0.1

• pytz 2020.4

• pyzmq 20.0.0

• CUDA 10.2

  on NVIDIA Titan RTX GPU. You'd better configure the environment as this.

Quick start

---

1. Clone the repo

git clone git@github.com:Badstu/CAKT.git

2. Install dependencies

pip install -r requirements.txt

3. Dataset

You can find the datasets at dataset folder, there are five datasets used in these projects.

The datasets are namely 'assist2009_updated', 'assist2015', 'assist2017', 'STATICS', 'synthetic'. You can change the dataset name parameter to run on different dataset.

4. Quick run

• Deep Knowledge Tracing

  You can run DKT model with main.py.

  cd DKT
  python main.py --dataset dataset_name    # change the dataset_name as you need
  

• Dynamic Key-Value Memory Networks for Knowledge Tracing

  You can run DKVMN model with main.py.

  cd DKVMN
  python main.py --dataset dataset_name    # change the dataset_name as you need
  

• Sequential Key-Value Memory Networks

  You can run SKVMN model with main.py.

  cd SKVMN
  python main.py --dataset dataset_name    # change the dataset_name as you need
  

• A Self-Attentive Model for Knowledge Tracing

  You can run SAKT model with main.py.

  cd SAKT
  python main.py --dataset dataset_name    # change the dataset_name as you need
  

• Exercise-aware Knowledge Tracing

  You can run EKT model with EKT_experiment.py.

  cd EKT
  python EKT_experiment.py dataset_name    # change the dataset_name as you need
  

• Convolutional Knowledge Tracing

  You can run CKT model with train.py.

  cd CKT
  python train.py dataset_name    # change the dataset_name as you need
  

  After the training process finished, You will see an instruction like this:

  The number 1607160632 is the trained model ID.

  To test the model, you can run with test.py.

  python test.py model_id dataset_name    # the model_id corresponds to the ID in the instruction after the training process
  										# the dataset_name here should accord with the training model
  

• Augmenting Knowledge Tracing by Considering Forgetting Behavior

  You can run DKT-F model with main.py.

  cd DKT_F
  python main.py --dataset dataset_name    # change the dataset_name as you need
  

• Context-Aware Attentive Knowledge Tracing

  You can run AKT model with main.py.

  cd AKT
  python main.py --dataset dataset_name    # change the dataset_name as you need