Document | DatasetInfo | PaperCollection | Experiment | RankingList

Important News!!!

We have migrated all code to PyEdmine, which significantly refactors and optimizes the framework while introducing a unified experimental pipeline for Knowledge Tracing (KT), Cognitive Diagnosis (CD), and Exercise Recommendation (ER). Additionally, PyEdmine open-sources all pre-trained model weights and parameters from the benchmark rankings.

Introduction

A library of algorithms for reproducing knowledge tracing, cognitive diagnosis, and exercise recommendation models.

Quick-Start

Prepare

1. Initialize project

   • Create file settings.json in the root directory.

   • Modify the environment configuration file settings.json

     {
       "LIB_PATH": ".../dlkt-main",  # Change to the project root path
       "FILE_MANAGER_ROOT": "any_dir"  # Any path used to store data and models
     }

   • Run set_up.py

     python set_up.py

2. Place the original files of the dataset in the corresponding directory (Please refer to Document (Section 1.3) for details)

3. Data Preprocessing: Run example/preprocess.py, for example

   python preprocess.py --dataset_name assist2009

Knowledge Tracing

1. Divide the dataset according to the specified experimental settings: Run example4knowledge_racing/prepare_dataset/akt_setting.py. For example,

   python akt_setting.py

   • For details on dataset partitioning, please refer to Document (Section 1.6)

2. Train model: Run the file under example/train. For example, train a DKT model

   python dkt.py

   • Regarding the meaning of parameters, please refer to Document (Section 2)

Cognitive Diagnosis

1. Divide the dataset according to the specified experimental settings: Run example4cognitive_diagnosis/prepare_dataset/akt_setting.py. For example,

   python ncd_setting.py

2. Train model: Run the file under example4cognitive_diagnosis/train. For example, train a NCD model

   python ncd.py

Exercise Recommendation

1. Divide the dataset according to the specified experimental settings: Run example4exercise_recommendation/prepare_dataset/kg4ex_setting.py. For example,

   python kg4ex_setting.py
   

2. Train or evaluate different model or method

   1. KG4EX

      • step 1, train a DKT model to get mlkc
      • step 2, train a DKT_KG4EX model to get pkc
      • step 3, run example4exercise_recommendation/kg4ex/get_mlkc_pkc.py
      • step 4, run example4exercise_recommendation/kg4ex/get_efr.py
      • step 5, run example4exercise_recommendation/kg4ex/get_triples.py
      • step 6, run example4exercise_recommendation/train/kg4ex.py

   2. EB-CF (Exercise-based collaborative filtering)

      • step1, change example4exercise_recommendation/eb_cf/load_data to get users' history data
      • step2, run example4exercise_recommendation/eb_cf/get_que_sim_mat.py to get questions' similarity matrix
      • step3, run example4exercise_recommendation/eb_cf/evaluate.py

   3. SB-CF (Student-based collaborative filtering)

      • Similar to EB-CF, run the code in example4exercise_recommendation/sb_cf

Contributing

Please let us know if you encounter a bug or have any suggestions by filing an issue

We welcome all contributions from bug fixes to new features and extensions.

We expect all contributions discussed in the issue tracker and going through PRs.

Contributors

• https://github.com/ZhijieXiong
• https://github.com/kingofpop625
• https://github.com/shshen-closer