Hershey, Pennsylvania
New York, New YorkBeijing, China
Sleep Quality Assessment Using Ensembles of Machine Learning and Deep Learning Models

Sleep Quality Assessment Using Ensembles of Machine Learning and Deep Learning Models

Kunaraj Kumarasamy (Loyola-ICAM College of Engineering and Technology (LICET), India), Maria Wenisch Sebastian (Loyola-ICAM College of Engineering and Technology (LICET), India), Robert Rajkumar Sakkariyas (Loyola-ICAM College of Engineering and Technology (LICET), India), and Vaithiyanathan Dhandapani (National Institute of Technology, Delhi, India)
Copyright: © 2021 |Pages: 10
DOI: 10.4018/978-1-7998-8018-9.ch011
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Estimating our sleep quality and state is essential to identify disorders or chronic ailments related to sleep patterns. The authors propose certain methods using machine learning and deep learning algorithms to assess the quality of sleep. The signals and hence the data taken from the wrist actigraphy and data accumulated through sleep research on postmenopausal women are used in this work. This data is preprocessed and used to score the sleep-wake pattern objectively and subjectively. With the availability of multi-ethnic data, both machine learning and deep learning models are created by rigorous training to avoid over-fitting and under-fitting. As the users of wearable active devices are increasing and proven to be a commercially successful model, this work is more relevant to multiple groups across ages.
Chapter Preview
Top

Machine Learning Classifiers

The features extracted from the actigraphy or smartwatch data are used as predictors and used to labels the sleep quality. The following classifiers are commonly used for these applications.

  • •

    Support Vector Machine

  • •

    Boosted decision trees

  • •

    Bagged decision trees

  • •

    K-nearest neighbor

Support Vector Machine (SVM)

SVM is a supervised learning-based machine learning classifier. Sumithra R, 2015 used SVM and KNN to classify skin lesions. Support Vector Machines make use of hyperplanes to classify data. The hyperplanes are decision boundaries that help distinguish between feature vectors of different objects. The best hyperplane is chosen with the help of support vectors which are points that are near the hyperplane. The best hyperplane is the one with the largest margin from the support vectors. The dimensionality of the hyperplane is determined by the number of feature vectors. We use the quadratic kernel function to define the hyperplane.

Boosted Decision Trees

A decision tree builds classification models in the form of a tree structure where the data is continuously split according to a certain parameter. Boosted decision trees are used to produce better predictive performance. It combines several decision trees in series and each decision tree aims at improving the error performance of the previous tree. The main drawback of boosted decision trees is that they take a longer time to train, owing to their series structure.

Bagged Decision Trees

A bagged decision tree is a type of ensemble technique that provides better accuracy when compared to decision tree classifiers. Bagged decision trees create subspaces of the training data and these subspaces are used to train independent decision trees. An aggregate of the probabilities of the individual decision trees is used to determine the final probability.

K-Nearest Neighbor

K-nearest neighbour is a simple machine learning algorithm suitable for classification that stores all available cases and classifies new cases based on a similarity measure. It classifies data points on the basis of its ‘k’ nearest neighbours. The value of k has to be suitably chosen to avoid over-fitting and achieve better training and validation error performance.

Complete Chapter List

Search this Book:
Reset