Factor analysis using PCA, SVD and bottleneck neural network (BNN) autoencoders using the Boston Housing dataset
Factor analysis is generally used for dimensionality reduction, which is achieved by identifying a low number of latent factors that are able to capture the variability of the data without too much loss of information.
The R script in this repository implements and compares different factor analysis techniques using data on the Boston real estate market:
- Principal component analysis
- Singular value decomposition
- Bottleneck neural networks
The different methods are compared based on how well they can discriminate between different house price classes.
The folder R contains the RStudio script for the implementation and the Rmarkdown file for generating the report.
A Docker image to run RStudio has been added to the repository to ensure reproducibility of results. See https://github.com/vettorefburana/Run-Rstudio-Server-from-Docker for instructions on how to run the Docker container.