Particle Swarm Optimization-Based and Grid Search Hyperparameter Tuning for Automatic Support Vector Regression Models in Cerebral Autoregulation Analysis
This package offers a comprehensive toolkit for the analysis and modeling of biological signal data specific to individual patients. It facilitates the training of Support Vector Regression (SVR) models to represent and predict cerebral autoregulation phenomena.
Features include:
- Utilizing k-fold cross-validation for training.
- Hyperparameter optimization through Particle Swarm Optimization (PSO) and Grid Search.
- Generation of various models trained with the following learning structures (both univariate and multivariate modeling):
- Finite Impulse Response (FIR)
- Nonlinear Finite Impulse Response (NFIR)
- AutoRegressive with eXogenous inputs (ARX)
- Nonlinear AutoRegressive with eXogenous inputs (NARX)
- An extended scoring filter to evaluate the quality of the autoregulation response when a patient is subjected to simulated pressure changes.
- Automatic selection for the autoregulation response.
# devtools::install_github("benjajorquera/PSOBrainModeler")- dplyr
- signal
- stats
- e1071
- utils
- pso
- progress
- tseries
- magrittr
- testthat
MIT License. See LICENSE file for more details.