A novel approach in oversampling algorithm for imbalanced data sets in the context of ordinal classification

Data sets are the backbone for data mining and knowledge engineering field. The class imbalance problem exists in many real-time data sets. In this paper we investigate the existing approaches for class imbalance problem in the context of classification and ordinal classification. In particular, this investigation extends the study of issues in ordinal classification with respect to the data set and it examines the current existing solutions. Of which Graph based approach is one of the preprocessing methods exists to deal with this problem. In this paper we propose to extend the existing graph based methodology to expand the probable area of synthetic patterns which provides an efficient technique for oversampling. Experimental results with data sets shows that, our proposed oversampling technique using collinear patterns yields better ordinal classification performance in terms of accuracy and sensitivity compared with currently existing solutions. This technique provides an effective and efficient solution for imbalanced complex data sets.