An Effective Approach Using Machine Learning for Fall Prediction

The ability to fall has now increased due to the usage of phones as well as a variety of distractions on the streets that has led to a major public concern. Fall is an activity that can be caused to carelessness or incoordination. It is also attributed to monitoring senior citizens who tend to have...

Full description

Saved in:
Bibliographic Details
Published in2023 9th International Conference on Smart Computing and Communications (ICSCC) pp. 315 - 320
Main Authors Vinora, A., Maheswari, P. Uma, Deborah, R.Nancy, Ajitha, E., Srinivasan, A., Sivakarthi, G.
Format Conference Proceeding
LanguageEnglish
Published IEEE 17.08.2023
Subjects
Backpropagation
Bayesian Regularization and Cross entropy
Entropy
Fall predicition
Gyroscope
Machine learning algorithms
Machine Learning(ML)
Performance evaluation
Prediction algorithms
Predictive models
Training
Online AccessGet full text
DOI10.1109/ICSCC59169.2023.10334969

Cover

More Information
Summary:The ability to fall has now increased due to the usage of phones as well as a variety of distractions on the streets that has led to a major public concern. Fall is an activity that can be caused to carelessness or incoordination. It is also attributed to monitoring senior citizens who tend to have a higher fall rate due to age-related illness, diseases, or poor visibility. It can be predicted for also soldiers for an unknown battle field based on their momentum. A dataset of 40000 rows have been obtained from Kaggle which was procured using a three-axis gyroscope that was deployed on a smartphone to collect data on falls during activities such as walking, jogging, standing, and sitting. A set of Machine Learning (ML) algorithms such as Scaled Conjugate Gradient, Bayesian Regularization, Levenberg-Marquardt, BFGS Quasi-Newton Backpropagation, Resilient Backpropagation, Conjugate Gradient Backpropagation, Conjugate Gradient Backpropagation With Polak-Ribiére Updates, One-Step Secant Backpropagation, Gradient Descent With Momentum And Adaptive Learning Rate Backpropagation and Gradient Descent With Momentum Backpropagation were deployed on the built model to check for fall prediction. Among them, Bayesian Regularization provided high accuracy rates considering cross entropy as a performance metric. This technology helps to prevent the fall of a person that can lead to an uneventful situation.
DOI:10.1109/ICSCC59169.2023.10334969