A genetic algorithm for feature selection in gait analysis

Smart Computational Approaches with Advanced Feature Selection Algorithms for Optimizing the Classification of Mobility Data in Health Informatics

“…[35] is a feature selection technique based on a maximum signal-to-noise ratio ranking combined with minimum correlation. Signal-to-noise(SNR) can be calculated by (1), in which the numerator is the difference of the average values of the same feature between two classes (e.g., PD and HE), and the denominator is the summation of the variance of the same feature within each class:…”

“…Feature selection and dimensionality reduction methods have been studied in the gait domain for a variety of applications, including activity recognition [13] and classification of different populations' gait [2,35]. These techniques have been used for identifying the best subset of features from the gait data collected using infrared markers [1,2], video images [32], ground reaction force mats [35], and wearable sensors [5,23]. Among the studies that used the data collected from wearable sensors to distinguish between the gait of healthy individuals and pathological gait, Caramia et al [5] investigates the various subsets of features on several classifiers' performance and reports that the subset of features built using PCA outperforms the other subsets of features made using the domain knowledge.…”

“…The result of a study by Rastegari et al indicates that their proposed Maximum Information Gain-Minimum Correlation feature selection method outperforms the Maximum Information Gain (MIG) algorithm in differentiating between PD patients and healthy age-matched elderlies using gait parameters [23]. Genetic algorithm (GA) and Maximum Signal-to-Noise Ratio-Minimum Correlation (MSNR-MC) feature selection methods have been used for feature selection in the gait domain [1,35] based on the data collected from infrared markers and ground reaction force mats, respectively.…”

Data Analytics in the Internet of Things: A Survey

“…It efficiently selects features from high dimensional data sets where exhaustive search is not feasible. Examples of IoT use-cases that have utilized genetic algorithm include intrusion detection [118], medical image feature extraction and selection [119], pattern recognition [120], building energy optimization [121], gait analysis [122], etc. Table 4.4 presents the purpose of dimensionality reduction techniques in IoT use-cases mentioned in this sub-section.…”