Professor Sifeng Liu selected to be one of the 10 shortlisted promising scientists in the MSCA 2017 Prizes
——A Summary on the MSCA project GS-A-DM-DS(629051)
The EU-funded Marie (Skłodowska) Curie Actions(MSCA) project: Grey Systems and Its Application to Data Mining and Decision Support (GS-A-DM-DS, 629051) was completed at the end of 2016. The research fellow, Professor Sifeng Liu has been selected to be one of the 10 shortlisted promising scientists in the MSCA 2017 Prizes
Combining the expertise of Chinese and European researchers, the GS-A-DM-DS project has led to significant advances in the reliability and applicability of data-mining algorithms based on grey systems theory, which enables valuable insights and predictions to be gained from incomplete data.
Grey systems models developed by the GS-A-DM-DS researchers were successfully used to help select a research and development team to work on China’s first domestic-built commercial passenger jet, the C919 that had its maiden flight in May. The grey system approach was chosen because China had no prior experience in building such an aircraft and therefore limited reference data.
A Chinese renewable energy company has also employed models advanced in the project to be able to predict when a gearbox on a wind turbine will fail, enabling it to prolong the life of its turbines by 36%, resulting in fewer shutdowns and saving tens of millions of euros.
“The C919 aircraft and the wind turbines are just two examples of how our models have been used in the real world to great effect,” says project manager Professor Yingjie Yang. “As our work focuses on the theory rather than the practical applications, we hope to show that grey systems models are a viable option for companies, governments and policymakers in areas with limited or poor quality data, such as socio-economic analysis, healthcare, climate change and complex R&D projects.”
The GS-A-DM-DS team proposed several new prediction and decision-making models to provide more reliable results in complex situations.
They formulated a set of criteria for grey model selection and calibration following systematic research. This will assist in making grey prediction and forecasting easily accessible to new users who have no prior knowledge in grey systems. The criteria will also help to promote the application of grey systems to data mining in Europe.
Project partners developed several grey models in order to achieve more accurate and reliable prediction and forecasting with “small data” and poor information. These will contribute to data mining operations that require high speed and reliability while reducing data requirements. They proposed the even difference model GM(1,1) (EDGM), the original difference model GM(1,1) (ODGM), self-memory grey model, and fractional order grey models, etc.
Researchers also developed decision-making models that were validated by simulation and real application case studies. Performance was superior to existing alternatives. These models will enable more realistic and reliable decision-making, and help ensure the uncertainty representation is more accessible for ordinary users.
One such example is the new grey clustering evaluation model, based on mixed possibility functions, which includes both end-point mixed possibility functions and centre-point mixed possibility functions. It’s easy to obtain the possibility functions and solve the evaluation problems of uncertain systems with poor information. They also studied the problems of multi-attribute intelligent grey target decision-making, and then constructed four kinds of uniform effect measure functions in view of the different decision-making objectives based on benefit type, cost type, and moderate type.
Accordingly, the various decision-making objectives which possess different meanings, dimensions, and/or nature from each other can now be transferred and measured to uniform effect. The critical value of a grey target is designed as the dividing point between positive and negative, which is defined as zero. The objective effect values were fully considered, and as a result, a new multi-attribute intelligent grey target decision-making model was proposed. Dealing with the decision-making dilemma of a comparison between the maximum components of two decision coefficient vectors is different from comparisons between the two integrated decision coefficient vectors themselves. Therefore, both the weight vector group of kernel clustering and weighted coefficient vectors of kernel clustering for decision-making were firstly defined. A novel two-stage decision-making model with the weight vector group of kernel clustering and weighted coefficient vector of kernel clustering for decision-making was then put forward. This method can effectively solve the decision-making dilemma and produce consistent results.
Over 30 research papers were presented in leading international academic journals and conferences, plus 3 books published by Springer, Science Press, and John Wiley & Sons, respectively. In addition, more than 20 visits, seminars and training courses were carried out in China and Europe. One of their research books on grey systems is identified as the No.1 top sited books in the pandect of natural science by China National Knowledge Infrastructure (CNKI) in 2017. The literatures of Prof. Sifeng Liu have been sited 22866 times in google scholar. His H-Index is 55, i10-index 329.
Today, the models are being used by oil companies to successfully predict where to drill new wells, saving considerable sums of money on speculative excavations; transport authorities are using the methods to estimate traffic on highways before they are built; and multinational brands are exploring the techniques to gain insights into how customers feel about their products.
As a result of the EU-funded Marie (Skłodowska) Curie Actions(MSCA) project, an international association on grey systems and uncertainty analysis was established comprising members from China, Europe and North America.
GS-A-DM-DS project has demonstrated the feasibility of grey systems in data mining and its great potential for use with limited and poor data. It will have a significant impact on the development of grey systems and data mining in China and Europe.
The GS-A-DM-DS team
(Sifeng Liu, Yingjie Yang, Zhigeng Fang, Lifeng Wu, Bo Zeng, Naiming Xie,
Chaoqing Yuan, Xiaojun Guo, Arjab Singh Khuman, Qiaoxing Li, Shuli Yan,
Jiefang Liu)
Institute for Grey System Studies, Nanjing University of Aeronautics and Astronautics
Centre for Computational Intelligence, De Montfort University
