Project Description
The term ‘Big Data’ is used to indicate massive amounts of complex data, be they structured or unstructured, real time or historical. Over the last few years, several important characteristics of Big Data have been highlighted: volume, velocity, and variety, have all received much attention.
However, an important aspect of Big Data is their usability, that is, are they good enough to be used for further analysis. The quality (or veracity) of Big Data may be degraded with the increase of dirty and imperfect data, which may originate from many sources, for instance, missing values, noise, systematic errors, and human errors.
Data pre-processing has been widely used in data mining tasks. In the context of Big Data this becomes extremely challenging. More sophisticated and effective approaches are urgently needed to deal with the quality issues surrounding Big Data and to improve their usability.
In this project we will investigate how machine learning (e.g. classification, deep learning, clustering) can be used to distinguish good data from bad ones and more importantly, to improve the usability of Big Data.
More specifically, we will investigate the following related topics:
(1) How can we extract salient features with respect to data quality so that these features can be used to train a machine learning model?
(2) How can we develop machine learning models specifically for evaluating data quality?
(3) How can we use machine learning to facilitate data imputation with an aim to improve the usability of Big Data.
(4) Can Bio-inspired computing and deep learning be used to develop such machine learning systems?
More:
https://www.findaphd.com/phds/project/machine-learning-approaches-for-improving-the-usability-of-big-data/?p81937
