Intelligent CPSS and its application to health care computing

Dayou Liu, Bo Yang, Shang Gao, Yungang Zhu, Yong Lai
2016 Science China Information Sciences  
The rapid development of cyber-physical systems (CPS) had a tremendous impact on human behavior and interactions. With human involvement, CPS has naturally evolved into cyberphysical social systems (CPSS) [1] . The top five technology breakthroughs of 2013 were closely related to CPSS and peripherals according to McKinsey's report [2]. CPSS takes full account of crowd intelligence, and requires effective integration of data, information and knowledge from diverse dimensions of computing,
more » ... ing, physical/social systems, etc., all with the purpose to integrate and analyze complex human behaviors. Thus, data/information/knowledge processing in CPSS based systems is challenging: not only do we need to solve massive, high-dimensional and noisy data problems, but also need to deal with further difficulties originating from data uncertainty, inconsistency, heterogeneity and spatio-temporal effects. With these challenges, the effective integration of distributed data mining, network data mining, data fusion, spatio-temporal reasoning and many other artificial intelligence methods is expected to play important roles. In this paper, we exemplify such integration in CPSS by using intelligent health-care computing in patient-oriented systems. Current data analysis methods mostly require the entire input data, not suitable for handling large-scale, distributed, decentralizd, and temporal data sets, which often characterize CPSS. Therefore, the integration of big and complex data sets is crucial. We believe that distributed data mining is one of the key areas, and suggest two basic strategies in this regard: divide-and-conquer and self-organization. For example, in applications with a large number of variables, which can be pair-wise correlated or independent to some degree or none, we can divide variables into different sets based on their associative relationships, then apply traditional techniques independently and merge the distributed results. Alternatively, one can take advantage of multi-agent systems, and build a selforganized, adaptive, and distributed data processing system, where agents in different CPSS components can gather and process local information and communicate through networks as well as collaborative protocols. By using multi-agent systems, locally collected and possibly hidden information can be effectively utilized in discovering global knowledge patterns. To solve complex problems in the real world, humans play a leading role. It has been controver-* Corresponding author (
doi:10.1007/s11432-016-5545-5 fatcat:yqwx6bkhkfcopg2g52joia3goe