image

Harnessing the ‘data deluge’ – Dr Norma Bargary awarded SFI Frontiers for the Future grant

Submitted on Wednesday, 18/11/2020

This month Minister Simon Harris announced 71 research grants through the SFI ‘Frontiers for the Future’ Programme. The research grants will impact areas such as spinal cord injury, novel materials, climate change, biodiversity in food production and waste, smart manufacturing, social connectivity, computer graphics, horse breeding, pharmaceutical manufacturing, and information security.

The Insight SFI Research Centre for Data Analytics is delighted to announce that Dr Norma Bargary has been awarded €467,569 for her project with Dr Andrew Simpkin:  Functional data Analysis for Sensor Technologies (FAST).

The FAST project will use mathematical and statistical models to describe the behaviour of, and relationships between, high-dimensional data arising from sensor technology. Data from sensors are collected in driver safety, medical imaging, e-health, manufacturing, ecology and the Internet-of-Things, and FAST will ensure the maximum amount of information is extracted from the data to aid robust decision-making.

Explaining the potential impact of the research, Dr Norma Bargery said: ‘Currently, there is much focus on machine learning with ‘big data’. Sensor data are regularly collected in medicine, sports science, and engineering, but are rarely utilised to their full capacity. Our algorithms, and freely available FAST R package, will provide a means to fully exploit the data in its richest form. FAST will provide a new framework for statisticians and data scientists to interpret, model and make predictions using sensor data, with the potential to provide lucrative insights to companies for decision-making and product development.’

The results of this project will ultimately impact diverse areas of societal and economic importance arising from the modelling of complex high-dimensional data found in the automotive industry, glucose monitoring in endocrinology, accelerometer data in sports science, Internet-of-Things for e-heath, and manufacturing, for example.

Wearable sensor technology is another key data source. Sales of wearable technology, such as FitBit, are expected to surge, with the market set to reach $150 billion by 2027. However, research from Price Waterhouse Cooper (PwC) indicates the effectiveness of wearable technology may decline, as there is no straightforward way to translate the considerable amount of data collected into actionable advice.

‘FAST will provide a toolbox of methods which ensures the maximum amount of information is extracted from data of this type for robust decision-making,’ according to Dr Bargery. ‘We will calibrate and test the FAST methodology against current state of-the-art modelling tools using datasets that measure groups of individuals at high frequency and from multiple sources simultaneously.

The medical technology sector in Ireland is recognised as one of five global emerging hubs, with eight of the world’s top ten med-tech companies based here.

The methodology developed in the FAST programme has the potential to impact many sectors; marketing, medical imaging, genomics, pharmaceutical, manufacturing; where measurement of numerous high dimensional sensor data streams is typical. For example in elite sports, data are being collected every day on groups of athletes with the aim of optimising performance and reducing risk of injury. However, sports scientists are suffering from ‘data deluge’ as expertise in data analytics lags behind the sensor data technology.

By creating a novel suite of methods and easy-to-use software FAST will enhance the capabilities of researchers to glean maximal inferences and make improved decisions using high-throughput sensor data. To ensure the widespread dissemination of this work to these broad sectors Drs Bargary and Simpkin will test the tools developed on open and diverse sensor datasets. The resulting software will be in the form of an open source R package combining data visualisation techniques, model fitting algorithms and example datasets so that researchers can apply these methods to their data.