Survey of Popular Linear Dimensionality Reduction Techniques

Abstract

Big Data analytics related solutions are one of the prime industrial focuses across all domains. In this digital era, the volumes of the data being generated by both machine and man are humongous. The key challenges are to store the data which requires more space and also to retrieve the data in an optimized way that saves time and money. The number of features for any given dataset is another problem statement. As in most real-life datasets, the number of features present is more and we are not sure, which features or dimensions are good enough to be considered. Visualization of the high dimensional data is also one of the most complex issues that impair meaningful insights since the visuals are not precise anymore due to feature explosion. In this paper, we have done a comprehensive survey on some of the most popular dimensionality reduction approaches broadly categorized under linear dimensionality reduction techniques. We have also done a critical comparative analysis focused on the utility of these algorithms. Finally concluded the survey with few notes on future work.KeywordsDimensionality reductionFeature extractionFeature reductionFactor analysisPrincipal Component AnalysisHigh dimensional data