AI xPress

AI and computer science research.

Back to Foragentis

Human-AI Interaction

A convolutional neural network classified university teaching quality with 80% to 92% accuracy in a test set

Human-AI Interaction November 21, 2026

A hand reaching to a car touchscreen

In a 2025 paper in Computers and Education Artificial Intelligence, researchers at Sanda University report that a convolutional neural network, a kind of deep learning algorithm, classified university teaching quality into four rating categories. The model reached 92% accuracy for “Excellent,” 88% for “Good,” 85% for “Average,” and 80% for “Poor” in a test set. The researchers combined student performance data, teacher evaluations, course content, and student feedback into one dataset. They encoded and standardized that data before training the model. The paper frames the approach as a step toward more objective teaching evaluation and personalized feedback.

What did the researchers do?

The researchers created a dataset from student performance data, teacher evaluations, course content, and student feedback. They encoded and standardized the data, then trained and validated a convolutional neural network on those inputs and analyzed the results on a test set.

What did the model achieve?

On the test set, the CNN classified “Excellent” teaching quality at 92% accuracy, “Good” at 88%, “Average” at 85%, and “Poor” at 80%. The authors also report high precision and recall for the classifications, though the abstract does not give those numeric values.

What does this mean for educational technology and pedagogy?

The authors argue that using several data sources can support a broader evaluation than traditional methods, which they describe as less rigorous and less comprehensive. They say the result provides a foundation for developing targeted teaching improvement strategies. The abstract does not say whether the researchers tested those strategies.

Frequently asked questions

What kind of model did the study use?

The study used a convolutional neural network, a type of deep learning model. The researchers trained it on data from student performance, teacher evaluations, course content, and student feedback.

How accurate was the model?

In a test set, the model classified “Excellent” teaching quality at 92% accuracy, “Good” at 88%, “Average” at 85%, and “Poor” at 80%.

Did the study test the approach in real classrooms?

The abstract reports results from a test set only. It does not describe a real classroom deployment.

Limits and what's next

The study evaluated university teaching quality on a dataset that combined student performance data, teacher evaluations, course content, and student feedback. The abstract does not state the number of students, instructors, courses, or institutions in that dataset. It also does not report how many observations fell into each rating category. This is one study based on a test set, and the abstract reports no independent replication or external validation. The abstract presents the findings as a foundation for developing targeted teaching improvement strategies. It does not report evidence from a real deployment.

Source: Gao Ying, “Deep learning-based strategies for evaluating and enhancing university teaching quality,” Computers and Education Artificial Intelligence, 2025.