On 1 December 2009, some 58 IT professionals and practitioners attended the afternoon seminar, Innovations in Governance and Architecture jointly organized by the Institute of Systems Science, NUS and Information Technology Standards Committee (ITSC).
Located in the tropics, Singapore suffers from prolonged rain during the Northeast Monsoon period.
The Singapore Meteorological Services Division is responsible for issuing warnings of upcoming prolonged rain. This is defined as rain that lasts more than 6 hours, with rainfall of over 75mm. Forecasters at the department make these predictions based on their experience, observation data, Numerical Weather Prediction models and simple rule of thumb.
But is there a better way to do this?
A team of five Master of Technology (Knowledge Engineering) students from ISS decided to use Artificial Intelligence techniques to develop a Knowledge Based System that would do a better job of forecasting.
The primary Artificial Intelligence technique used is Case Based Reasoning. The team also utilised Data Mining techniques for exploring various new parameters that could potentially be of important in the forecasting of prolonged heavy rain.
Essentially, the system would make available the experience of past years to all forecasters, who would not have to spend time to accumulate these experiences. It also provides a more exhaustive and systematic way of recollecting past events, compared to human memory.
The project was challenging for the team. Only one member of the team had any basic understanding of the weather system. The team also had to wade through seven years of past weather data.
Adding to the challenge, the team bravely decided to build its own Neural Networks.
The Neural Networks system created by the students has the capability for cascading recurrent networks with hidden and output layer feedback and processing time-sequenced data. In addition, the series data can be weighed linearly or exponentially so that more recent events are assigned a higher value. Data visualisation is employed in real-time to enable the team to view the progress step by step. The system also offers comparative plots and histograms, weights distribution plots and more. As a result of these challenges, the project took over 2,000 man-hours to complete.
However, the result was worth all the effort. The final prototype provides a glimpse into the potential of using Artificial Intelligence and Soft Computing techniques for this domain of weather prediction.
The project is a first in Singapore. But more work lies ahead, stressed Charles Pang, the Chief, Knowledge Engineering Programme for ISS and the supervisor on this project. He said, "The system we've developed is a prototype. More work and research needs to be done and it will be several more prototype iterations before we actually can tell whether the system is successful and deployable."
For their work on this project, Ang Cheng Hai, Rajiv Juneja, Lim Wee Long, Vanitha Ramaswamy and Wong Mung Jih, Michael were named winners of the Best Master Of Technology (Knowledge Engineering) Project in 2008.
Winners of the Best Master of Technology (Knowledge Engineering) Project. (standing from left) Michael Wong, Rajiv Juneja, Lim Wee Long, (seated from left) Ang Chieng Hai and Vanitha Ramaswamy
For their graduation project, a team of ISS' Master Of Technology (Software Engineering) students asked themselves this question: How do you get data search tools for genetics research to go from primitive to phenomenal?
There are few realms more exciting or more sophisticated today than the field of genetic research. The same, however, cannot be said of research tools for the domain.
In the post-genomic era, humankind has advanced in a very short time from having little information about genes to having an immense amount of data on them. Getting to the data, however, means plowing through hundreds of databases. Over 500 genetic databases exist in the public sphere - each of which has to be accessed individually.
For ISS' Master of Technology (Software Engineering) students Ashok Arockiaraj Savarimuthu, Petrus Cornelis Johannes Bruin, Chia Shue Ching, Pang Wei Mun, Renganayaki Ramasamy and Jacob Schondorf, the answer was simple: Develop a system that, given a biological record, will search several databases for information that may be of interest to a researcher studying the record. Specifically, the team aimed at creating a system that will aid biologists to identify protein sequences that have good probability to serve as vaccine targets.
But working without a prior background in biological science, the team struggled. Related the team, "The WADE system is a very unique and innovative system; there isn't any system similar in functional characteristics. When developing unique systems, it is difficult for the user to provide well defined and complete requirements. And so our team had to do a great amount of research on the biomedical domain, on the characteristics of the 3rd party tools, as well as technical research on various charting tools that could meet our requirements."
Eventually the team designed and developed a framework architecture and the bio database driver's interfaces. The system is built on a 3-tier software architecture developed on the Rational Rose Unified Process (RUP) and based on three underlying technologies. They are Presentation Tier (a web-based system utilising Google Web Toolkit (GWT) technology), Application Tier (which uses Java Enterprise Edition servers) and Data Tier (which uses MySQL Database).
The final system, called Web based Aggregation and Display of Epitope (WADE), uses advanced graphical display technology for an easy and intuitive user interface. In its backend, WADE is a powerful platform that can interrogate many computational algorithms in a robust manner.
The system has huge potential: The system would allow researchers to drastically cut the amount of time they spend on research.
Recounted the team, "The biomedical staffs at the local university's faculty of medicine were delighted and excited with our system when we demonstrated it to them for the first time. Instantly they made several requests to enhance the system."
Commended project supervisor Olivo Miotto of the team, "They have met and exceeded my expectations, and produced a great piece of software. The system works very well, appears very professional and is easy to use. What is even more exciting is that its initial use has already prompted discussion on a number of further developments. We are at the start of an exciting journey." Mr Miotto has since left ISS and is today the Senior Informatics Fellow at Oxford University's Centre for Genomics and Global Health.
For their work on this project, the six ISS students were named winners of the Best Master Of Technology (Software Engineering) Project in 2008.