ACEMS Session on ‘Big Data, Big Models and New Insights’
The ACEMS special session on Big Data, Big Models and New Insights will host three exciting talks on big data.
“Big Data and Visual Analytics as a Tool to Gain New Insights”, Tomasz Bednarz, ACEMS & Queensland University of Technology
Big data is now endemic in business, industry, government, environmental management, medical science, social research and many other fields. One of the challenges is determining how to effectively model and analyse these data, find new correlations and discover new patterns. This talk on Big Data, Big Models and New Insights will discuss the challenges of building and operating big data platforms, ways to unify big data frameworks for various scientific investigations and computations, and emphasize the global privacy concerns with such systems. It will also introduce new projects connecting analytics and visualisation, that are currently undertaken at the Queensland University of Technology node of ACEMS. Tomasz will also introduce ACEMS in general.
“Multi-scale two-directional two-dimensional principal component analysis for large array biomedical signal classification”, Hong-Bo Xie, ACEMS & Queensland University of Technology
The steady improvement of biomedical recording techniques has increasingly permitted the registration of a high number of channels. However, conventional feature reduction methods such as principal component analysis and linear discriminant analysis are not applicable to high-dimensional recordings due to the curse of dimensionality and small sample size problem. We present a multi-scale two-directional two-dimensional principal component analysis (MS2D2PCA) method for the efficient and effective extraction of essential feature information from large array biomedical signal recordings. Multi-scale matrices constructed in the first step incorporate the spatial correlation and physiological characteristics of sub-band signals among channels. In the second step, the two-directional two-dimensional principal component analysis operates on the multi-scale matrices to reduce the dimension, rather than vectors in conventional PCA. To demonstrate the efficiency and effectiveness of the proposed method, results are presented from an experiment to classify 20 hand movements using 89-channel electromyographic signals recorded in stroke survivors.
“Symbolic data techniques for analysing time-dependent histogram data”, Xin Zhang and Scott A. Sisson, ACEMS & University of New South Wales
Symbolic data analysis provides a way to simplify complex and high-dimensional datasets into composite objects such as hypercubes and histograms. This is useful for both visualisation purposes, and also in principle for data analysis, although techniques for the latter are underdeveloped. In this presentation we will introduce the main ideas behind symbolic data analysis, and present some initial results in analysing time-dependent histogram data. This work is motivated by the need to analyse the evolution of airborne particle pollution data over time, which is directly recorded in histogram form.