Multi-objective Optimization for Dimension Reduction for Large Datasets

In recent advancement of computational techniques, there is an exponential increase in amount of data. Learning on such large amount of data is a major area of concern with application of machine learning algorithms. Therefore, it is considered to be a complicated task to handle and perform computation on such large, complex, and heterogenous dataset. In this paper, a brief discussion about different dimension reduction or feature selection algorithms is given. A brief review about contribution of researchers for designing feature selection algorithms for large dataset is given. By analyzing exisitng problems, this paper is motivated to design a hybrid, robust, flexible, and dynamic feature selection model for classification of large datasets. For this, multi-objective optimized feature selection is proposed with an objective to minimize the error rate and execution time as well maximize accuracy of problem and to generate solution with high probability.