engleski

Dimensionality Reduction for Non-linear Local Segmentation and Profiling

Growth of data collection on consumer behavior has created a huge explosion in dimensionality of the collected feature space. As a consequence of the expanding feature space, the curse of the dimensionality produces sparsity of data creating issues for machine learning as the amount of data to obtain statistically sound results increases exponentially. Dimension projection techniques are used to combat the problem by reducing and compressing the feature space to a more manageable size using transformations of the original space into a lower dimensional structure. The key goal of feature projection is the extraction of key behavioral information with the reduction of redundant or unnecessary information in the original feature space. The expected result of the feature projection is a lower-dimensional space that still adequately describes original data with "minimal" loss of information.

Machine Learning ; Dimensionality Reduction ; Segmentation ; Profiling

nije evidentirano