Συγκριτική μελέτη γραφοθεωρητικών αλγορίθμων ημι-επιτηρούμενης μηχανικής μάθησης σε προβλήματα ταξινόμησης με μεγάλη ταξική ανισορροπία
Comparative study of graph-based semi-supervised machine learning algorithms on classification problems with extreme class imbalance
Αλεξανδροπούλου, Χαρίκλεια K.
This thesis, entitled “Comparative Study of Graph-Based Semi-Supervised Machine Learning Algorithms on Classification Problems with Extreme Class Imbalance”, is a description, analysis and implementation of graph-based Semi-Supervised Learning algorithms. Initially we attempt to bibliographic approach the field of machine learning with partial supervision. Then, the algorithms are analyzed theoretically and applied to data classification of classes with large class imbalance. An applied correction method is also used. This method uses the prior knowledge of the classes. Finally, we propose a new correction method, based on the correct information which is obtained from the previous one and this information is being further developed.