Classification of self efficacy beliefs of information technology teachers about Scratch software by machine learning and deep learning methods

With the development of technology, the skills that are expected from individuals have also changed. Computational thinking is one of the most important skills of 21st century.A qualified coding education is needed in order to upskill the individuals for computational thinking. Concordantly, the Ministry of National Education has included coding education in the scope of the information technologies curriculum. In this context, this training is given by using Scratch which is one of the basic block-based coding tools. Therefore, it is extremely important how competent Information Technologies teachers consider themselves about coding tools. In this study, it is aimed to classify the self-efficacy beliefs of Information Technologies teachers regarding the use of Scratch by using classical machine learning and deep learning methods.T-SECT scale which was previously developed was used as a data collection tool. This scale consists of 39 Likert type items. The data set was created by using 192 samples consisting of Information Technology teachers and 39 attributes. In order to eliminate the problem of unbalanced data due to the limited number of samples, the data was amplified by the SMOTE method and the number of samples was increased to 262. The data set was transferred into WEKA software and classical machine learning methods, Google Colab platform and Convolutional Neural Network (CNN) methods were used on it. In line with this purpose, classification success was calculated with Zero-R, J48, Random Forest(RF), K-Star, Multilayer-Perceptron(MLP), NaivesBayes, SMO, Logistic, IBK and Random Tree methods. The classical machine learning methods with the highest classification performance on the data set were found as SMO, MLP, IBK, J48 and RF. Classification which is made with CNN received better results than classical machine learning methods. Self-efficacy beliefs of Information Technologies teachers were successfully classified using CNN with an accuracy rate of 99.30%.