AUTHOR=Chen Zhonelue , Li Gen , Gao Chao , Tan Yuyan , Liu Jun , Zhao Jin , Ling Yun , Yu Xiaoliu , Ren Kang , Chen Shengdi 

TITLE=Prediction of Freezing of Gait in Parkinson’s Disease Using a Random Forest Model Based on an Orthogonal Experimental Design: A Pilot Study

JOURNAL=Frontiers in Human Neuroscience

VOLUME=Volume 15 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2021.636414

DOI=10.3389/fnhum.2021.636414

ISSN=1662-5161

ABSTRACT=Purpose: The purpose of this study was to introduce an orthogonal experimental design (OED) to optimize the of feature extraction parameters for predicting the freezing of gait (FoG) in Parkinson’s disease (PD).
Methods: A random forest model was developed to predict FoG in PD by identifying possible precursor signs of FoG (preFoG) in PD using acceleration signals and angular velocity signals. An OED was introduced to optimize the hyperparameters of feature extraction, and their main effects and interactions on model performance were analyzed.
Results: We simultaneously obtained optimized parameters and analyzed the main effects and interaction among hyperparameters of feature extraction hyperparameters. The false positive rate, hit rate and mean predicted time were 27%, 68%, and 2.99 s, respectively.
Conclusion: The OED was very effective for investigating the main effects and interactions between feature extraction parameters. It was also beneficial for optimizing the feature extraction parameters to enhance the performance of the FoG prediction model for PD.