In our Project, the small dataset classification is the main issue; because, the insufficient data in the dataset will not lead to a robust classification performance. So, to improve the accuracy of small dataset analysis, we propose a new approach, Attribute construction process i.e.) Extending the attribute information. It is based on kernel based techniques. These techniques used to transform the input space into a higher dimensional feature space for extending the attribute information, using the classification oriented fuzzy membership algorithm. The Attribute Construction process based on two approaches; class possibility attributes construction, Synthetic attribute construction. For constructing the attribute, find the area overlap function, if it is low means; add the class possibility values as the new attributes. When the overlap area is high means, instead of adding the new attributes, to construct new attributes, for substituting the original ones, to enhance the classification accuracy. Compared to previous methods, like Principal Component Analysis (PCA), Kernel Principal Component Analysis (KPCA), & Kernel Independent Component Analysis (KICA), the proposed data attribute construction procedure is to extend the data information of small datasets to improve the classification performance.