Open AccessFeature Engineering: Towards Identification of Symptom Clusters of Mental Disorders
Author(s) -
Mariam A. Sheha,
Mai S. Mabrouk,
Amr Sharawy
Publication year - 2022
Publication title -
ieee access
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.587
H-Index - 127
ISSN - 2169-3536
DOI - 10.1109/access.2022.3232075
Subject(s) - aerospace , bioengineering , communication, networking and broadcast technologies , components, circuits, devices and systems , computing and processing , engineered materials, dielectrics and plasmas , engineering profession , fields, waves and electromagnetics , general topics for engineers , geoscience , nuclear engineering , photonics and electrooptics , power, energy and industry applications , robotics and control systems , signal processing and analysis , transportation
Mental and behavioral problems primarily affect the mind and brain, altering emotions and cognition. Society is heavily burdened by mental disorders, particularly if they are not identified or treated early. Despite the significant influence of mental health on people’s ability to function in many areas of life, biomedical engineering research remains remote. This study engages feature engineering and machine learning to find the significant characteristics that margin between (healthy controls and depression), and (healthy controls and schizophrenia) performing computer-aided diagnosis systems for mental health diagnosis. This research extracted twenty-seven features based on several well-known scientific functions, such as entropy, correlation, and energy, to identify two of the five main mental symptom clusters (psychomotor and disorganization). The most expressive features were selected according to four techniques based on the wrapper and filter methods, then examined using five classification techniques. Sixty different experimental models were executed for each dataset separately, based on various forms of data representation, feature selection techniques, and classification methods. In addition, fifteen multi-label experiments were performed to challenge the system’s proficiency in studying each group versus the other two groups (one verses rest). The results indicate that the features used are auspicious in determining participants’ mental states and can effectively employ actigraphy data to distinguish between different mental conditions. Experimental models based on daily time series segmentation are the best to apply, considering the number of observations and the classification performance.