Information Technology involves in both cross-sectional and longitudinal research activities. IT@JCU researches cutting-edge computer solutions for various domain-specific problems through artificial intelligence, machine learning, data mining, mobile computing, and cryptographic approaches.
It has been a limiting factor to gain reliable data in many fisheries and aquaculture management applications. This project tries to solve bottlenecks in fisheries and aquaculture industries through machine learning and deep learning approaches utilising varying levels of human supervision. Application of deep learning will provide significant efficiencies in counting and measuring abalone and will remove handling induced stress and errors.
Monitoring tropical coastal fish population status is complicated by species diversity and diversity of common fish names. This research seeks to reduce training costs and improve survey efficiency by developing an app using AI and deep learning to automatically identify and measure reef fish species. Augmented Reality is used to automate and enhance data collection, and deep learning will be utilised at the back end for fish type classification. This will significantly increase reporting efficiency by removing the need to identify and measure fish, process the data and run analysis. While the app will be developed for fish type classification, it has applicability in other object measurement and classification.
As we move to data-rich and computation-rich environments, deep learning approaches including Convolutional Neural Networks and Recurrent Neural Networks have been wisely used for numerous applications and domains including ecology, underwater monitoring, medical imagery, aquaculture, engineering, biology, etc. This is to investigate algorithmic improvements of deep learning, and to apply state-of-the-art deep learning algorithms to solve domain specific problems.
Immersive technologies such as virtual reality and augmented reality have been used along with serious games and simulations to provide realistic and interactive experience to various target users. This is of particular use if a target is rare and not easily observed in the real world. One typical example is tree kangaroo that is difficult observe in the wild, and there is extremely limited chance for direct or indirect interaction with it. Virtual re-birth of these rare animals naturally brings awareness of the species and also indirect experience.
Multi-party computation (MPC) protocols provide a general model for secure computation of arbitrary function whose arguments (inputs) are held by a group of participants. A secure MPC protocol enables a set of mutually distrusting participants, each with their own private input, to compute a function such that at the end of the protocol, all participants learn the correct value, while the confidentiality of the private inputs is maintained.
Understanding human movements and behavioural patterns is a hot research topic. Tourism is one of big industries in Australia, and it is of great importance to understand where tourists go, what they do, where they spent time and money, etc. This research investigates various data mining techniques to find sequential, periodic, and associative movement and behavioural patterns for informed and data-driven decision making. For instance, 90% of Asian tourists coming to Cairns airport visit Kuranda and Tjapukai cultural park in the first day, and drive to Port Douglas for the second day before they go to the Great Barrier Reef in the last day. This sequential pattern could help tourism industry to make informed decisions.
Staff scheduling and rostering problem has become increasingly important as business becomes more service oriented and cost conscious in a global environment. The development of an optimised roster model for a specific workforce has not been easily achieved due to the complexity of rostering a specialised workforce and the difficulty of configuring resources to achieve both the cost saving and employees satisfaction. This study is about exploration the use of various optimisation algorithms and implementation of an automatic roster system framework to optimise utilisation of existing workforce resources. The system implemented provides an artificially intelligent solution to optimisation-modelling of workforce logistics.
There has been rapid growth of text data in the context of different web-based applications such as social media. Therefore, this study explores a number of different methods and algorithms which can effectively process a wide variety of text applications. The most recent project Dr. Joanne Lee has been involved is an application of text mining on the medical literature as an empirical basis for curriculum mapping and assessment of medical training.
Product reviews are constantly posted online on websites like Amazon. Natural language processing aims at automatic summarisation and recommendation using written text. This includes identifying positive or negative sentiments and the aspect causing them. Conversation agents such as chat-bots can also be trained to discuss a certain topic. This is achieved by converting each word to a vector representation such that semantically similar words are close to each other in the vector space. Sentiment analysis could be applied to various textual formats such as medical prescription, emails, web documents, etc.
Discharge measurements are vital in water resource management and flood management. Compared to the current standard discharge measurement using ADCP, an IoT solution for discharge measurement has the advantages of: low-cost equipment, no staff required in the field, fully automated process and the ability to provide continuous near-real-time discharge data – even to remote sites. This revolutionary solution means that we will be able to monitor the discharge of waterways on a large-scale with continuous data monitoring and near real-time alerts that we never had before. This technology will significantly improve water resource management and flood event management, with greatly reduced health and safety risks.
We can predict the emotional state of a person from his facial expression. Facial action units such as a 'smile' allow us to accurately determine the presence of emotions. Convolutional neural networks have shown high accuracy in classifying images. Recently, adversarial models are also being used for generating face emotions. This can be used for classification of YouTube reviews in new languages such as Spanish.
There has been a large volume of trajectories collected due to the adoption and growth of sensor-enabled mobile phones and GPS tracking technologies. A trajectory is a series of moving entity data, commonly in a format such as (longitude, latitude, timestamp), and it can represent any object’s movement such as human traces, vehicle tracks, animals with ear-tags or sensors, hurricanes, goods and products with RFID tags, etc. This is to investigate trajectory related data mining approaches such as simplification, stop/move detection, sequential patterns, periodic patterns, clustering, and path recommendation.
This research investigates the gap between university and industry for Information Technology graduates in regional and metropolitan Australia, and evaluates what could be done by both universities and industry to bridge this gap. This could include preparing IT graduates effectively using more industry-relevant education methods.