Research in the design and implementation of the SMASH (Semantic Mining of Activity, Social, and Health data) system will address a critical need for data mining tools to help understanding the influence of healthcare social networks, such as YesiWell, on sustained weight loss where the data are multi-dimensional, temporal, semantically heterogeneous, and very sensitive. System design and implementation will rest on five specific aims.
The first aim i s to develop a novel data mining and statistical learning approach to understand key factors that enable spread of healthy behaviors in a social network (Aim 1). We propose to develop a formal and expressive Semantic Web ontology for the concepts used in describing the semantic features of healthcare data and social networks. We will then bridge the domain knowledge in healthcare and social networks with formal mappings across those ontological concepts (Aim 2). Next, we propose novel recommendation approaches building on top of the influence modeling and prediction. In addition, we will develop methods to utilize the recommendation as a means to better organize the social network such that the adoption of optimal health behaviors in the network can spread quickly and sustainably (Aim 3). To protect the privacy of human subjects during the data mining process for social network and health data, we consider the enforcement of differential privacy through a privacy preserving analysis layer. We will develop novel solutions to preserve differential privacy for mining dynamic health data and social activities of human subjects (Aim 4). To support this research, we will develop a web- accessible portal so that other researchers with little training i data mining will have shared access to data mining tools, ontologies, and social network analysis results (Aim 5). At the end of this project, data resources, tools, ontologies, and technologies will be made available to the larger research community. This work is an inter-disciplinary collaboration among the PI, Dejing Dou, Co-I Daniel Lowd, both experts in data mining and machine learning, and Jessica Greene, an expert in health policy, at the University of Oregon, Brigitte Piniewski MD, the lead of YesiWell, at PeaceHealth Laboratories, Ruoming Jin, an expert in complex network mining, at Kent State University, Xintao Wu, an expert in privacy preserving mining, at the University of North Carolina at Charlotte, David Kil, the previous Chief Scientist at SKT Americas and program manager of YesiWell, and the founder of HealthMantic, and Junfeng Sun, a mathematical statistician at the NIH and an expert in design of clinical trials.
Frequent social contacts are effective to sustained weight loss. The goal of this research is to develop advanced data mining tools, formal ontologies, privacy preserving methods, and web-portal to help the understanding of the influence of social networks on healthcare outcomes, in particular sustained weight loss. It will help other researchers in healthcare with little training in data mining to have shared access to the data mining tools, ontologies, and social network analysis results.