In my essay I chose a data-mining problem, more precisely the multi-class learning, which I describe in more detail in the first section. After this description I go through three recent papers about the problem then introduce some data-mining terms and algorithms related to my topic.
During my work I used two different datasets. In the first one there were 1000 pictures' grey-scale feature descriptions. My first task was to pre-process these into a structure, on which I could execute classification algorithms later. In the second database I had the relations of 25000 pictures and 94 tags recorded. In this case instead of pre-processing the images the first step was discovering the relationships between the categories. In my essay first I present my work and results on the first database, then go through the work I did on the second database. During my work on the first database I used LIBSVM and Weka, while in case of the second database I used my own program to build a tree from the categories based on the relations discovered between them. Later on I used this tree to classify the test database's pictures.