This is an implementation of the HELP-DKT model, described in the following paper: HELP-DKT: an Interpretable Cognitive Model of How Students Learn Programming Based on Deep Knowledge Tracing (https://www.nature.com/articles/s41598-022-07956-0).

• Yu Liang: [email protected]
• Tianhao Peng: [email protected]
• Meng Wang

State Key Laboratory of Software Development Environment Admire Group, School of Computer Science and Engineering, Beihang University

The original Python code files are compressed in the path: ./Data/Original_Codes.zip. Please unzip it.

unzip ./Data/Original_Codes.zip

Our dataset includes 9119 source codes collected from a Python Programming Introductory course hosted in a MOOC platform for learning a variety of programming languages.

Each Python file in the dataset directory represents a student's submission, and the name of the file is organized as follows:

1. the result of submission('b' for buggy, 'c' for correct)
2. the challenge number of submission(e.g. '362', '371')
3. the student ID
4. the number of student's submissions on one challenge

Consider one submission 'b_362_27176_2'. The example represents:

1. 'b': buggy, the submission is error
2. '362': the challenge number is 362
3. '27173': the student ID is 27173
4. '2': the Python file is the student's second submission for the challenge

Another example 'c_449_27303_5':

1. 'c': correct, the submission is correct
2. '449': the challenge number is 449
3. '27303': the student ID is 27303
4. '5': the Python file is the student's 5th submission for the challenge

In our paper, we use 'C-1', 'C-2', ... , 'C-6' to represent the challenges instead of '362', '449',... , '472' for clarity. The following table shows the comparison relationship:

see Code_Program_Embeddings/README.md

N.B. if you do not want to reproduce generating the program vector embeddings, you can start here to run the HELP-DKT model using the embedding results in Program_Vector_Embeddings.CSV.

see Code_HELP_DKT/README.md

@article{liang_help-dkt_2022,
	title = {{HELP}-{DKT}: an interpretable cognitive model of how students learn programming based on deep knowledge tracing},
	volume = {12},
	url = {https://doi.org/10.1038/s41598-022-07956-0},
	doi = {10.1038/s41598-022-07956-0},
	number = {1},
	journal = {Scientific Reports},
	author = {Liang, Yu and Peng, Tianhao and Pu, Yanjun and Wu, Wenjun},
	month = mar,
	year = {2022},
	pages = {4012},
}