A small selection of the many projects underway in the DSKD Lab are summarised below.
Predicting human behaviour on video
Software that can intelligently interpret human behaviour may sound like Big Brother, but the DSKD team will make it a reality sooner than we think. The benefits of being able to use video surveillance to fight crime and combat terrorism are clear and prescient, but the technology has wider applications that could save lives on a daily basis, such as predicting traffic accidents or helping trucks safely merge on to a freeway. For the past few years Prof Dacheng Tao has been developing advanced probabilistic methods and Bayesian multi-linear analysis frameworks that can be applied to surveillance footage to warn of impending accidents, fight crime and predict human behaviour. More recently, he has extended this research to partner with Toyota to help develop technology that will make vehicles smarter and safer.
Location-aware pattern mining
The uptake of GPS and advances in radar sensor technology have given rise to a new wealth of spatial and information to add to our expanding knowledge of population behaviour. Prof Chengqi Zhang and Dr Ling Chen are exploring how this information might be used to diagnose weather patterns, assist markets and detect crime by applying location data to improve traditional pattern mining techniques. At its core their research not only discovers which events frequently co-occur but also where those occurrences take place and this work is giving rise to an entirely new field of research in pattern mining.
Computation vs communication costs
Because of its ability to model complex relationships in the real world, big graph processing is becoming increasingly important in industry and academia. Dr Lu Qin, recipient of a 2014 Chancellor’s Postdoctoral Fellowship, is working to find scalable solutions for big graph processing using MapReduce that allow a wide range of potential design algorithms while maintaining a feasible trade-off between computation and communication costs.
Enterprise pricing models
Empowering large enterprises with the capability to set real-time, dynamic pricing is one of the ways in which the DSKD lab is helping Australian businesses develop world-leading technology. Using frameworks based on deep pattern mining, the team is addressing critical and currently under-developed knowledge discovery and data mining issues with the aim of transforming existing policy-driven business intelligence models into deep pattern-driven, real-time, dynamic pricing tools.