The use of video clickstream data to predict university students' test performance: A comprehensive educational data...

Area of Science:

• Educational Data Mining
• Learning Analytics

Background:

• Video clickstream behaviors (e.g., pauses, forwards, backwards) are common in online learning.
• These behaviors present opportunities for educational data mining and learning analytics.

Purpose of the Study:

• To investigate the predictive relationship between video clickstream behaviors and student test performance.
• To explore how specific clickstream patterns correlate with academic success.

Main Methods:

• Two experiments were conducted: an exploratory study with 22 students and a conclusive study with 16 students.
• Data mining techniques and repeated measures were employed to analyze video clickstream data and test performance.
• Statistical analysis included correlation and predictive modeling (Root Mean Squared Error Percentage).

Main Results:

• A positive correlation was found between total clicks and test performance.
• Students with high click counts, slow backward speed, or more pauses/backwards performed better.
• Video clickstream data accurately predicted test performance (15-20% RMSE).
• Mean backward speed, number of pauses, and backward frequency were key predictors.

Conclusions:

• Video clickstream behaviors can predict student test performance with notable accuracy.
• Specific behaviors like slow backward navigation and pauses indicate higher academic achievement.
• Findings can help identify at-risk students and inform the design of video-based learning materials.