Python, C, C++, Matlab
Python, C, C++, Matlab
Embedded System, Sensor Design
Multiple Sclerosis is a complicated disease with varying symptoms which causes neurological disability in people with little chance of a full cure. The goal of this work is to ensure a healthy and normal lifestyle for MS patients through better symptoms and better behavior. This is necessary to improve the healthcare outcomes of MS patients so that they may be able to function better at home and work with effective, comprehensive MS care outside the clinic. The knowledge gained from such exploration may also impact the design of future devices to coach MS patients based on their symptoms to perform activities and improve the quality of their life. MS is a complex, non-curable disease, and medication alone can not help modify or slow the disease course. Performance of physical activities by MS patients may also cause changes in their symptoms and in turn impact the health outcome, which is not studied yet. We want to use behavior modeling to explore how healthcare outcomes and MS symptoms change over time based on the physical activities being performed by MS patients. For this approach, accelerometer-based continuous daily activity data along with the self-reported momentary assessment of symptoms (pain level, fatigue level) may help us to correlate the contextual state and actions performed in those states by MS patients to know how these chronic symptoms affect people and if any interventions based on behavior can affect their healthcare outcomes.
One of the goals of recommender systems is to enable consumers to choose the right product that is personalized for their needs. For example, the right skin care products applied at the right schedule and as part of a personalized plan can greatly help consumers prevent and address bad skin days. However, existing product recommender systems recommend products based on high level similarity between consumers (e.g., their demographics, skin, and products they use), but not based on their product usage schedules or plans. Although such systems could lead to good product selection, they are unable to provide deep insights about consumers’ behaviors or coach the consumers about how to schedule or plan proper usage of the selected products personalized for their particular needs. This is because consumers are different (e.g., they have different kinds of skin and require different skin care products); what works for one person may or may not work for another. We propose to develop a system that leverages consumer/product interaction data to infer consumer behaviors, reflect on their usage of different products, and understand schedules and plans what work for them. To do this, we propose to apply our existing methodology for modeling human routine behaviors to create a computational model of consumer routine behavior. Such a computational model will capture people’s preference for different products and the outcomes of product use (e.g., good skin/bad skin days) over time, and automatically learn their product use plans and schedules.