Clustering Analysis of Online Teaching Cases and Evaluation of Teaching Results

Shuyuan Bi, Wei LiuInternational Journal of Emerging Technologies in Learning (iJET) 2023 · DOI

Abstract

To facilitate front-line teachers in providing more efficient cases for classroom teaching activities within limited teaching time, it is necessary to conduct clustering analysis of online teaching cases. Existing clustering studies on text-based teaching resources often focus on the representation and fusion of different views but ignore the high-dimensional sparsity of text data. This paper presents a design flow for online teaching with case teaching as the core and proposes a text multi-view clustering algorithm based on case subject alignment. By utilizing a clustering model structure based on subject alignment and grey system theory, the study evaluates online classroom teaching results before and after the application of the algorithm. Experimental results verify the effectiveness of the proposed approach in organizing resources and improving pedagogical outcomes.

Introduction

In the era of curriculum reform and STEAM education, case-based teaching has emerged as a highly effective method for engaging students with real-world scenarios. However, the preparation phase is often time-consuming for educators. While online repositories offer a vast sea of teaching cases, these resources are frequently disorganized and difficult to navigate. This research addresses the need for automated, intelligent organization of these resources to help teachers find the most relevant materials efficiently.

The Challenge of Sparse Multi-View Data

Traditional text-clustering methods face two primary hurdles when dealing with online teaching cases:

  1. High-Dimensional Sparsity: Teaching cases contain thousands of potential features (words), but any single case only utilizes a small subset, making it difficult for standard algorithms to identify meaningful patterns.
  2. Multi-View Inconsistency: A single teaching case may be described through various "views," such as a summary, learning objectives, or student feedback. Standard models often fail to reconcile the structural inconsistencies between these different perspectives, leading to inaccurate categorization.

Proposed Methodology: Subject Alignment Clustering

The core of this research is a Text Multi-View Clustering Algorithm based on Case Subject Alignment. This model leverages a Variational Autoencoder (VAE) framework to learn the underlying latent structure of teaching cases through deep variational inference.

Architecture of the text multi-view clustering model based on subject alignment

As shown in the architecture above, the model employs an attention-based alignment operator using Key, Query, and Value mechanisms. This allows the system to "align" different views of a case, ensuring that the global subject representation remains consistent even when individual descriptions vary. This approach prioritizes informative views (like case summaries) over less descriptive metadata.

Pedagogical Integration: From Deconstruction to Design

Beyond the technical clustering, the paper outlines a comprehensive flow for integrating these organized cases into the classroom. The process begins with case deconstruction, breaking down resources into logical structures and knowledge content.

Design flow of online teaching with case teaching as the core

This deconstruction allows for the classification of knowledge points into declarative, procedural, and strategic categories. These categories then inform the online teaching design, which progresses from self-learning and testing to collaborative discussion and final teaching evaluations. This structured approach ensures that the clustered cases are utilized to their full pedagogical potential.

Experimental Results and Evaluation

The proposed algorithm was benchmarked against established models including DeepWalk, DNGR, and SGNS using custom datasets from English online teaching scenarios.

  • Clustering Accuracy: The proposed algorithm achieved a perplexity score of 362 in Sample Set 1, significantly outperforming DeepWalk (692) and DNGR (714). Lower perplexity indicates a better fit for the data.
  • Teaching Effectiveness: Using Grey Relational Analysis and multiple linear regression, the researchers found that the quality and type of clustered cases explained 83.6% of the variance in teaching effectiveness ($R^2 = 0.836$). Factors such as the "enlightenment" and "authenticity" of the cases were found to have statistically significant positive impacts on student outcomes.

Conclusion

By addressing the high-dimensional sparsity of text data through a subject alignment mechanism, this research provides a robust tool for educational resource management. The model not only categorizes cases with high accuracy but also demonstrates a clear correlation between organized resource delivery and improved classroom results. Future work will explore the application of this semantic alignment across more diverse and technical disciplines.

Summary updated

Abstract

To facilitate front-line teachers to provide more efficient cases for classroom teaching activities in limited teaching time, it is necessary to conduct clustering analysis of online teaching cases. The existing clustering study on text-based teaching resources focuses on the representation and fusion of different views of texts and the discovery of consistent clustering allocation, and ignores the high-dimensional sparsity of text data. To this end, this paper studies clustering analysis of online teaching cases and evaluation of teaching results. This paper gives the design flow of online teaching with case teaching as the core and proposes a text multi-view clustering algorithm based on case subject alignment, and presents a clustering model structure based on subject alignment. Based on the grey system theory, this paper compares online classroom teaching results before and after teachers use clustering algorithm to process teaching cases. Finally, the experimental setup is given, and the effectiveness of the proposed algorithm is verified by experiment.
BibTeX
@article{bi2023clustering,
  author = {Shuyuan Bi and Wei Liu},
  title = {{Clustering Analysis of Online Teaching Cases and Evaluation of Teaching Results}},
  journal = {International Journal of Emerging Technologies in Learning (iJET)},
  year = {2023},
  doi = {10.3991/ijet.v18i03.38055},
}
Figure