Geriatrics and Parkinson Disease Gait Classification based on Machine Learning Approaches and A Correlation Network Model
Advisor Information
Hesham Ali
Location
UNO Criss Library, Room 231
Presentation Type
Oral Presentation
Start Date
2-3-2018 1:00 PM
End Date
2-3-2018 1:15 PM
Abstract
Healthcare is moving rapidly from the long-standing reactive treatment approach to the early detection and preventative era. However, to fully embrace this trend, new approaches need to be developed. A step in this direction is to explore how to leverage data collected from wearable sensors to help in assessing health levels. This would pave the way for continuously monitoring individuals, which, in turn, lead to helping physicians diagnose diseases in the early stages. However, a major missing piece in moving forward with this concept is the lack of a sophisticated data analytics model. In this study, we use our proposed correlation network model as well as various machine learning approaches, including Random Forest and Adaboost to classify three groups of individuals, including patients with Parkinson Disease, Geriatrics, and Healthy elderlies by utilizing various gait parameters associated with groups under study. First, we extract features from raw acceleration data. Input features are pre-processed using two types of normalization techniques. Then the effectiveness of feature vectors to identify each group of individuals is evaluated for both correlation network model and machine learning approaches. Results show that using normalized features increases the both models' accuracy.
Geriatrics and Parkinson Disease Gait Classification based on Machine Learning Approaches and A Correlation Network Model
UNO Criss Library, Room 231
Healthcare is moving rapidly from the long-standing reactive treatment approach to the early detection and preventative era. However, to fully embrace this trend, new approaches need to be developed. A step in this direction is to explore how to leverage data collected from wearable sensors to help in assessing health levels. This would pave the way for continuously monitoring individuals, which, in turn, lead to helping physicians diagnose diseases in the early stages. However, a major missing piece in moving forward with this concept is the lack of a sophisticated data analytics model. In this study, we use our proposed correlation network model as well as various machine learning approaches, including Random Forest and Adaboost to classify three groups of individuals, including patients with Parkinson Disease, Geriatrics, and Healthy elderlies by utilizing various gait parameters associated with groups under study. First, we extract features from raw acceleration data. Input features are pre-processed using two types of normalization techniques. Then the effectiveness of feature vectors to identify each group of individuals is evaluated for both correlation network model and machine learning approaches. Results show that using normalized features increases the both models' accuracy.