Plug-and-Play EEG-Based Student Confusion Classification in Massive Online Open Courses (AIED 2023)

https://link.springer.com/chapter/10.1007/978-3-031-36272-9_57

Codes for running analysis on students' confusion levels via EEG-based deep neural network classsification. The leave-one-subject-and-video-out (LOSVO) paradigm is introduced here as a means to accurately simulate a plug-and-play implementation of the proposed classification framework into real-world classrooms.

Results Overview

Non-Normalized EEG Data

Videos	1	2	3	4	5	6	7	8	9	10	Average
Subject 1	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00
2	100.00	72.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	97.00	86.90
3	100.00	91.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	83.00	97.40
4	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	86.00	98.60
5	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00
6	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00
7	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	0.00	100.00	90.00
8	100.00	100.00	100.00	64.00	100.00	100.00	100.00	100.00	0.00	100.00	86.40
9	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00
10	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	82.00	98.20
Average	100.00	96.30	100.00	96.40	100.00	100.00	100.00	90.00	80.00	94.80	95.75

Normalized EEG Data

Videos	1	2	3	4	5	6	7	8	9	10	Average
Subject 1	62.00	100.00	100.00	48.00	65.00	100.00	100.00	95.00	100.00	100.00	87.00
2	95.00	100.00	100.00	100.00	97.00	100.00	100.00	88.00	76.00	87.00	94.30
3	97.00	95.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	99.20
4	100.00	100.00	100.00	97.00	100.00	100.00	100.00	100.00	100.00	86.00	99.70
5	35.00	100.00	100.00	100.00	100.00	82.00	100.00	100.00	88.00	100.00	90.50
6	95.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	99.50
7	70.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	94.00	96.40
8	100.00	100.00	100.00	96.00	100.00	42.00	100.00	100.00	100.00	100.00	93.80
9	100.00	100.00	100.00	46.00	100.00	100.00	100.00	100.00	100.00	100.00	94.60
10	13.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	95.00	90.80
Average	76.70	99.50	100.00	88.70	96.20	92.40	100.00	98.30	96.40	97.60	94.58

Resources

Dataset: Link

EEGNet Framework: Link

Dependencies

Run

Subject-and-Video-Independent

Download the input folder from the dataset link (requires a Kaggle account) and put in the main folder directory along with the code. Otherwise, simply download the data from this repository.

usage: python LOSVO_DCN.py [-vid VID] [-subj SUBJ] [--normalize]

Trains and Evaluates a binary classifier for students' confusion levels based on the EEGNet framework under the leave-one-subject-and-video-out (LOSVO) paradigm.

Required Arguments:
    -vid VID                            Set the video ID which data will be removed from training the classifier
    -subj SUBJ                          Set the subject whose data will be removed from training the classifier

Optional Arguments:
    --normalize                         Default is set to false, utilize this argument to normalize the EEG data

For example, to obtain the accuracy of subject 5 for video 5 with normalization on EEG data, the following code can be run

python LOSVO_DCN.py -vid 5 -subj 5 --normalize

Or without normalization:

python LOSVO_DCN.py -vid 5 -subj 5

---

Alternatively, to obtain the LOSVO accuracy for all subjects for each individual video, run LOSVO_DCN_all.py, with the option to normalize the EEG data using the normalize argument --normalize.

python LOSVO_DCN_all.py --normalize

---

Subject-Independent and Video-Independent

Likewise, subject-independent and video-independent models are available for testing.

Subject-independent:

python LOSO_DCN.py -subj SUBJ --normalize

Video-independent:

python LOVO_DCN.py -vid VID --normalize