Analysis of Time Series for the Intelligent Support to Early Mobilization with Machine Learning

Identifier:  TFM:356

Handle: https://hdl.handle.net/20.500.11797/TFM356

Authors:  Martin Colville, Alexander David

Abstract:
The purpose of this study is the analysis of machine learning techniques based on big data analysis and artificial neural networks to support early mobilization techniques on patients in an Intensive Care Unit. More specifically, the goal is to recognize the movement of patients to monitor their evolution overtime, dealing with time series data obtained from accelerometers. This document outlines an implementation where multiple algorithms (Support Vector Machines, Gaussian Naïve Bayes, Decision Trees and Multilayer Perceptron) are used in conjunction with time series to train models study their accuracy. Furthermore, a validation dataset is created with off-the-shelf devices to obtain new motion predictions. The project is aimed to establish two main facts: the effectivity of machine learning techniques when dealing with time series and their forecasting and the possibility of the integration of these for widespread applications.