Predicting learner's performance through video sequences viewing behavior analysis using educational data-mining.

Area of Science:

• Educational Technology
• Learning Analytics
• Data Mining

Background:

• Understanding learner behavior in online educational videos is crucial for improving learning outcomes.
• Traditional analysis often focuses on click types, neglecting the context of pedagogical sequences.
• Educational videos are increasingly used, necessitating methods to analyze learner engagement within them.

Purpose of the Study:

• To analyze learner interactions with pedagogical sequences in educational videos.
• To predict learner performance (pass/fail) based on their navigation patterns within video courses.
• To provide instructors with insights into learner behavior for timely intervention.

Main Methods:

• Segmenting educational videos into pedagogical sequences.
• Collecting and classifying learner click data based on sequence context.
• Applying educational data mining techniques, specifically K-nearest Neighbors (kNN) and Multilayer Perceptron (MLP) algorithms.
• Utilizing classification to predict learner performance.

Main Results:

• Learner performance in video courses can be predicted based on their interaction with pedagogical sequences.
• The kNN classifier achieved the highest accuracy at 65.07%.
• A significant correlation was observed between video sequence viewing behavior and learning performance.

Conclusions:

• Learner navigation patterns within pedagogical sequences of educational videos are indicative of their academic success.
• The developed method offers a valuable tool for instructors to monitor student engagement and identify potential learning difficulties early.
• This approach facilitates proactive instructional support, potentially enhancing overall learner achievement in video-based courses.