Efficient High Dimensional Data Classification

The present century is the century of big data. Recent advancements in technology have made huge amounts of data available. The trend today is towards not only collecting more patterns but rather to collect a larger number of variables that describe each pattern. The automatic and systematic collection of finer details of each pattern has led to high dimensional data. The classical classification methods are not designed to cope with this kind of explosive growth of the dimensionality of individual patterns. The demand for large number of patterns grows exponentially with the dimensionality of the feature space. To overcome this problem either we should reduce the dimensionality or increase the size of the training set by adding some artificially generated training patterns to the training set. Dimensionality reduction methods have been well studied. In this chapter we present the various methods employed in pattern synthesis and its effect on classification performance of both kNN (K-nearest neighbor) and SVM (support vector machine) classifiers.