Predicting pre-service teachers' computational thinking skills using machine learning classifiers.

Area of Science:

• Education
• Computer Science
• Educational Technology

Background:

• Computational thinking (CT) skills are crucial for educators, yet CT training effectiveness remains inconsistent.
• Identifying key predictors of CT skills is essential for optimizing teacher training programs.

Purpose of the Study:

• To develop an online CT training environment for pre-service teachers.
• To compare the predictive performance of four machine learning algorithms for CT skill classification.
• To identify significant predictors of CT skills in pre-service teachers.

Main Methods:

• Developed an online computational thinking training environment.
• Utilized supervised machine learning algorithms (Decision Tree, K-Nearest Neighbors, Logistic Regression, Naive Bayes).
• Employed log data and survey data for model training and validation.

Main Results:

• Decision Tree algorithm demonstrated superior performance in classifying pre-service teachers' CT skills compared to other models.
• Time spent on CT training emerged as a primary predictor.
• Prior CT skills and perceived difficulty of learning content were also significant predictors.

Conclusions:

• Machine learning models, particularly Decision Tree, can effectively predict CT skills in pre-service teachers.
• Training duration, existing CT proficiency, and learner perception of difficulty are critical factors influencing CT skill development.