Machine learning model to predict the level of adaptability in online education

Today, machine learning modeling methods used in many different sectors have shown high performance in areas where they are used. The use of machine learning, which is used in different sectors, has become widespread with the rapid development of technology. The methods used in different sectors have seen and solved many problems. With the development of technology, the Education sector has also been affected by this. The use of distance learning or online learning has increased with the Corona period. However, it seems that many examples of online learning used as a learning method are not successful. The aim of this study is to apply an example method by presenting a strategy for using the online learning method for this existing problem. Thus, it is aimed to contribute to ongoing studies in the field of technology. For this purpose, 2000 lines of data with demographic and economic characteristics were examined and underwent some pre-processing and visualization processes. In the study, after performing some data processing, models were established and the student's adaptation level to online learning was determined. While the level of online compliance in the data set is selected as the dependent variable, other variables are evaluated holistically and each of them affects the level of online compliance. Then, to create a model on these data, the attribute was determined by correlation and the models were created again with the selected algorithms. In these models, training and data sets were tested as 60%-40%, 70%-30%, 80%-20% and 90%-10%, respectively. Then, the performance of the models, f1-score, precision and recall values were compared. As a result of the tests carried out with the created models, the algorithm with the highest success level was determined to be the model used as 70% training set and 30% data set in the Decision Trees algorithm created after correlation. As a result of the study, the success performance of the model was found to be 0.85, f1-score 0.85, sharpness 0.77 and recall 0.88. According to the model results, the group with the highest level of online adaptation is male students with 71 people. Among male students, the compatibility level of university students aged 21-25 studying in private schools is high. When we further customize this data, the group with a low economic status, using a computer as a means of accessing the course, wi-fi as an internet connection, and 4G as a network type has the highest compatibility. When the results obtained are examined, people's demographic and economic characteristics affect online learning compatibility. It is envisaged that taking this situation into consideration, working at regular intervals and preparing the learning design using this parameter will affect the success of people.