Determination of secondary school students achievements with machine learning methods and an application

Machine Learning; It is applied in many fields, including education, in order to draw meaningful results by analyzing the data and to develop prediction models using input data. In this thesis, it is aimed to predict the year-end weighted grade point averages with machine learning methods by using the socioeconomic, demographic characteristics and course grades data of secondary school students and to determine the qualities that affect the success the most. The data set used in the study was obtained from a secondary school in Tuzla, Istanbul. By choosing the features on the data set, the variables that have the most impact on academic achievement have been determined and these attributes have been used in modeling. With the aim of classifying the target attribute; 7 machine learning algorithms, K-Nearest Neighbor, Decision Trees, Random Forest, Support Vector Machines, Artificial Neural Networks, Logistic Regression and Naive Bayes, have been applied and the performance performances of the models have been compared. According to the model performance evaluation results, the most successful model has been determined as the Random Forest algorithm. The qualifications of the students' academic background (past years grade point average, Turkish lesson first term average and Mathematic lesson first term average) have been determined as the attributes that affect the success the most. Absenteeism has been identified as another important factor affecting success. Although demographic and socioeconomic characteristics do not have much effect on academic success, it has been observed that they contribute to prediction success. At the end of the study, a sample system has been developed for predicting the academic success of secondary school students using the Random Forest model. The sample system can be reached at https://model-tahmin.herokuapp.com. It is thought that the developed system can be used by education administrators.