Akademik Veri Yönetim Sistemi
ACTA INFOLOGICA, cilt.7, sa.2, ss.360-383, 2023 (ESCI, TRDizin)
Designing a prediction method with machine learning algorithms and increasing the prediction success is one of the most important research areas and aims of today. Models designed using classification algorithms are frequently used especially in problem types that require prediction. In this study, real life data is used to answer the question of which problem type should be included in the Information Technology Service Management (ITSM) system. An important step in the search for a solution is to examine the dataset with regularization methods. Experimental results have been obtained to establish the overfitting or underfitting balance of the dataset with L1 and L2 regularization methods. While the Root-Mean-Square Error (RMSE) value was approximately 0.13 in the regression model without regularization, this value was found to be approximately 0.083 after L1 regularization.With the regularized dataset, new results were obtained using Artificial Neural Network (ANN), Logistic Regression (LR), Support Vector Machine (SVM) classifier algorithms. SVM algorithm was the most successful model with a performance of approximately 0.73. It is followed by LR and ANN respectively. Accuracy, Precision, Recall and F1Score were used as evaluation metrics. It is seen that the use of regularization methods, especially in the preparation of real-life data for use in machine learning or other artificial intelligence research, will contribute to increasing the success level of the model.