Attributed Multiplex Heterogeneous Network (AMHEN) Representation Learning for Students and Courses in Higher Education
Evaluating sources of course information and models of representation on a variety of institutional prediction tasks, Weijie Jiang and Zachary Pardos. In Proceedings of The 13th International Conference on Educational Data Mining (EDM 2020).
This repo includes code for training the proposed Attributed Multiplex Heterogeneous Network (AMHEN) to learn student and course (grade) embeddings, which was inspired by the paper:
- Cen, Y., Zou, X., Zhang, J., Yang, H., Zhou, J. and Tang, J., 2019, July. Representation learning for attributed multiplex heterogeneous network. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (pp. 1358-1368).
Preprocessed files required to train the AMHEN in our paper on an enrollment dataset:
-
train.txt: Each line represents an edge (grade) connecting a student node and a course node, which contains three tokens <edge_type> where each token should be a number.
-
node_type.txt: Each line contains two tokens <node_type>, where <node_type> should be consistent with the meta-path schema in the training command, i.e., --schema node_type_1-node_type_2-...-node_type_k-node_type_1. (Note that the first node type in the schema should equals to the last node type.) For enrollment data, only two types of nodes are supported, i.e., student node and course node.
-
feature.npy (optional, only feature of courses is supported currently): Each row represents features of a course, the number of which corresponds to course number.
- python3
- pytorch
- install other dependencies by
- pip install -r requirements.txt
-
For data without course attributes:
- python src/main.py
-
For data with course attributes:
- python src_attri/main.py
Please do not hesitate to contact us (jiangwj[at]berkeley[dot]edu) if you have any questions. We appreciate your support and citation if you find this work is useful.
@inproceedings{jiang2020evaluating,
title={Evaluating Sources of Course Information and Models of Representation on a Variety of Institutional Prediction Tasks.},
author={Jiang, Weijie and Pardos, Zachary A},
journal={International Conference on Educational Data Mining},
year={2020},
pages = {115--125},
publisher={ERIC}
}