Naslov
Gait speed prediction based on walking parameters using MLPRegressor

Autori
Radulović, Dejan ; Negovanović, Dino

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
Proceedings of the International Scientific Student Conference RI-STEM-2021 / Lorencin, Ivan ; Baressi Šegota, Sandi - Rijeka, 2021, 13-18

ISBN
978-953-8246-22-7

Skup
International Student Scientific Conference (Ri-STEM 2021)

Mjesto i datum
Rijeka, Hrvatska, 10.06.2021. - 11.06.2021

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Artificial intelligence ; Machine Learning ; MLPRegressor ; Multilayer Perceptron

Sažetak
The aim of this paper is to regress the walking speed from all other parameters in given dataset. Given dataset consists of the parameters that are divided into four groups: Basic Parameters, Temporary Parameters, Spatial Parameters and Height Parameters. Based on the given data, the goal is to train the neural network to predict gait speed from other features. Many Python libraries have been used, and in addition to pandas and statistics, the most important are sklearn.neural_network and sklearn.model_selection. The neural network architecture used is a multilayer perceptron (MLP), and the regression quality is evaluated using R2 and RMSE metrics. A GridSearchCV was implemented, which was used to optimize ie. tune hyperparameters and the results are presented in tables that contain true, predicted values, mean squaed errors and standard deviations. The main conclusion is that by increasing the number of hidden layers that define the structure of the artificial neural network, the training takes longer but also reduces the error, ie. the difference between the true and predicted values.

Izvorni jezik
Engleski

Znanstvena područja
Elektrotehnika, Temeljne tehničke znanosti, Interdisciplinarne tehničke znanosti

POVEZANOST RADA