Attributes and Cases Selection for Social Data Classification

The current paper presents an effective method to improve the classification of social data, by selecting relevant cases (objects) and attributes (features). This is accomplished using a hybrid approach that combines metaheuristic algorithms and Rough Set Theory. When selecting some relevant attributes and cases of the training data of the Nearest Neighbor classifier, this model has been found to be more efficient in the correct discrimination of objects. Experimental results show that applying hybrid algorithms for training set preprocessing contributes to increment the desired efficiency and robustness of the classifier model over social data.