Optimizing generalization ability of machine learning / Mohamed Abdullah Mohamed ; Supervised Hegazy M. Zaher , Naglaa R. Saeid

Thesis (Ph.D.) - Cairo University - Institute Studies and Statistical Research - Department of Operations Research

One of the main challenges in machine learning classification is enhancing the generalization ability. In particular, it is building a classification model that has ability to generalize is a significance challenge for researchers. This thesis tackles the problem of optimizing generalization ability of machine learning through selecting the appropriate structure of machine learning based on the main features that represent the data. The main hypothesis in this thesis is that selecting the appropriate structure of machine learning will minimize the true error of classification; in other words, maximizing the generalization ability of machine learning. The thesis displays the literature review of the generalization ability field in addition introducing classification of the related work based on the school that considered this issue. The thesis presents four issues of machine learning in the field, tackled in novel way. Firstly, optimizing generalization ability of machine learning based on individual learning by sequential hybridization between version space and the multi- criteria technique,TOPSIS,algorithm. The main idea of this approach is using the multi- criteria technique, TOPSIS, to rank the hypotheses according to its generalization ability, and then select the highest rank

Issued also as CD