Dimensionality Reduction Big Data using Information Theoretic Learning
Dimensionality Reduction, Information Theoretic Learning, Big Data, Smart Cities
Currently the various areas of work involving information technology are generating unprecedented amount of data through its applications and scenarios indicating further growth. In smart cities of the application of this work focuses, data systems contribute substantially to the increase of the volume of data. These systems involve different subsystems targeted to help public administration for housing and living of the population of urban centers. Traffic control, garbage collection, public transportation, health and safety are examples of areas where systems are developed for the purpose of smarties cities. Each system alone already generates a large amount of data enough to hinder the decision making process when they are integrated is practically impossible to interpret them and generate results for the solution of problems of everyday life. This large set of data, known as Big Data, has a wide variety of types and makes it quite difficult to be processed using traditional approaches; hence reducing the dimensionality of the data becomes almost essential to the decision-making process. The purpose of Dimensionality reduction is to preserve as much information contained in a high-dimension in its representation in low dimension, in other words, the idea is to reduce the number of high dimension attributes to a low dimension, preferably two or three dimensions, thus preserving the maximum information and knowledge contained in the data. In the bibliography there are several techniques for reducing the dimensionality, the spread of this work is the use of techniques of Information Theoretic Learning (ITL) in implementing this reduction, thus improving performance achieved by other techniques.