Fuzzy Rough Set Based Classification of Semi-supervised Data
In several classification applications, obtaining labelled training instances is costly or difficult. While the feature values of observations can be relatively easy to collect, sufficient resources and expert knowledge are required to (manually) assign these elements to their correct classes. This phenomenon is addressed by the introduction of semi-supervised classification, in which a prediction model is derived from a training set consisting of both labelled and unlabelled data. Information in both the labelled and unlabelled parts of the training set can be used in the classification process.