Home | News | Jobs | Research Topics | People | Projects | Publications | Resources | Blogs | Students | Seminars | Visiting | Contact 


Distributed Solar Prediction

ML Members: Bob Williamson, Edwin Bonilla, Justin Domke, Nick Engerer, Stephen Gould, Mark Reid, Christfried Webers, Xinhua Zhang

This project aims to predict the solar energy production of photovoltaic cells on rooftops in a larger city more accurately, more quickly, and less expensively than current technologies. Forecasting the solar energy production over a time of minutes to hours is important in order to stabilise the energy grid, decrease the operational cost, and ultimately achieve a high penetration of distributed solar energy production. We employ low cost image and radiation sensors to collect real-time data, develop machine learning and computer vision techniques, and devise quantitative measures of performance to create a model for forecasting the solar energy production within a suburb-size region in Canberra.

Making Machine Learning Transparent

Making Machine Learning Transparent

ML Members: Fang Chen, Jianlong Zhou, M. Asif Khawaja, Zhidong Li, Yang Wang, Ronnie Taib, Jinjun Sun

An ML algorithm is as a “black-box” for non-ML experts, which limits its impact on more general widespread applications. NICTA’s research focuses on making the ML process understandable and usable by end users through revealing internal real-time status update of ML models with meaningful presentations. The research helps to formulate guidelines/standards for the user interaction design of ML-based applications. As a result, ML results from transparent ML are interpreted more easily, more meaningfully, and more convincingly for end users.


Advanced Data Analytics in Transport

ML Members: Fang Chen, Chen Cai, Hoang Nguyen, Ronnie Taib, Scott Sanner, Aditya Menon, Jinjun Sun, Tao Wen, Raymond Wong 

NICTA researchers are researching in machine learning and data analytical algorithms in solving real-world transport problems. Data generated from road traffic are abundant in volume and diverse in variety, creating the challenge of learning and making sense of them. NICTA’s research uses probabilistic graphical models, non-parametric Bayesian modelling, and causal inference to address this challenge, leading to solutions for predicting traffic patterns, providing decision support, ensuring road safety, validating performance metrics, and creating social benefits. The NICTA solutions can benefit a large scale of organisations, including traffic management authorities, public transport systems, private delivery companies for freight and people, and other relevant service providers.



OpinionWatch and EventWatch

ML Members:  Scott Sanner, Paul Rivera

Analysts, researchers and consumers can be overwhelmed with content from reports, news, blogs and tweets. Typical tasks required may be combining written opinions of multiple experts, analysing news, or assessing public consultations. We provide standard document search and browse facilities as a canvas with visualisation and co-occurrence of key words, phrases and names.  Our unique capability is the integrated interactive visualisation of document topics, sentiment and timeline.This lets the analyst explore and understand the collection in terms of discovered common underlying themes, their time and opinions.

Click on the following links for more information on OpinionWatch or EventWatch:

Decision Support for Incident Management (DSIM)

Decision Support for Incident Management (DSIM)

ML Members: Fang Chen, Eric Choi, Julien Epps, Ronnie Taib, Yang Wang

NICTA and the NSW Transport Management Centre are collaborating to improve traffic incident handling through early prediction of clearance time and other crucial business factors. Using advanced Machine Learning techniques on historical incident handling data, patterns of normal and abnormal actions can be identified and used to provide decision support to operators, and efficient communication to the road users.

Data driven pipe failure prediction

Data Driven Pipe Failure Prediction

ML Members : Fang Chen, Eric Choi, Julien Epps, Ronnie Taib, Yang Wang, Zhidong Li, Bang Zhang

More accurate pipe failure prediction and preventive condition assessment for Australia's urban water infrastructure has the potential to save over $400M annually on reactive repairs and maintenance, as well as better prioritise capital spending and minimise disruption to water supplies, transport and traffic. NICTA's Machine Learning techniques use existing and historical data to predict future failures in water networks with the aim of delivering cost savings to Australia's water utilities.



ML Members: Leif HanlenHanna Suominen

NICTA's Canberra Research Laboratory has a focus on the use of Machine Learning to utilise the resources available over eHealth media.  With projects in the Northern Territory related to tele-health, the analysis of hospital admission information to derive admission predictions, this team is focused on advanced data mining and statistical modelling methods.

Geothermal Project

Geothermal Exploration

ML Members: Fabio Ramos, Edwin Bonilla, Simon O’Callaghan, Alistair Reid

NICTA’s robust research capabilities in machine learning and in the increasingly significant area of big data analytics will be used to locate and characterise geothermal energy sources deep beneath the surface of the Earth.

Big Data

Big Data Analytics for Enterprise

ML Members: Tiberio Caetano, Dimitri Semenovich, Rukshan Batuwitage

The Enterprise is in a prime position to leverage big data analytics: they have large, rich data sets, and many business problems that can be solved using data. The data assets that can be exploited in the Enterprise are sources such as CRMs, call logs, transactions, and web logs. Sometimes these data assets can be leveraged directly from existing database systems or data warehouses, and other times they must be accessed from the relevant source systems.
The results produced by a big data analytics system solving a real business problem must be integrated back into the organisation. This turns otherwise simply interesting results into actionable insights. In an enterprise context, many systems can potentially be the sink for results. For example, results could be populated back into CRMs to provide access to per-customer insights, or results could be used to drive targeting marketing or adverting campaigns, either through traditional or digital channels.


Driver Mental State Monitoring

ML Members: Ronnie Taib

The Driver Mental State Monitoring (DMSM) project is a research collaboration between NICTA and Germany's Fraunhofer IESE, focusing on the real-time assessment of drivers' mental states through behavioural and physiological inputs, in view of improving user experience, safety and training for the automotive industry and end-users.



ML Members: Fang Chen, Julien Epps, Benjamin Itzstein, Phu Ngoc Le, Lipeng Li, Ronnie Taib, Yang Wang

In many industries, human agents or operators are required to cope with high volumes of complex information, often under pressure. This results in high mental demand (and consequently high cognitive load), which often leads to errors and omissions, improper behaviour, and customer dissatisfaction. In the call centre industry, this ultimately leads to high staff turnover and associated costs . In military, infrastructure control and financial industries, the impact can be even greater.
NICTA has created the first Cognitive Load Monitoring (CLM) System based on the automatic analysis of speech. Grounded in a strong theoretical foundation, the patent-pending CLM engine (BrainGauge®) quantifies the cognitive load experienced by an operator from live voice recordings in real-time. BrainGauge® is unique because it directly measures mental load, and it is not based on keyword-spotting, which does not accurately reflect the ability to manage multiple tasks

Past Projects