On your device

---

Prior starting, make sure to have admin rights on your device. While this is not a must-have it will make your life way easier.

A. Basics

1. Install R

• https://cran.r-project.org/bin/windows/base/
  • We developed the code in \(\texttt{R}\) 3.4.1. While backward compatibility is not guaranteed, newer \(\texttt{R}\) releases should properly work.

2. Install R Studio (R IDE)

• https://www.rstudio.com/products/rstudio/download/

3. Data Manipulation

install.packages('tidyverse')

4. General ML package

install.packages('caret')

5. Others packages in support

install.packages(c('rmarkdown',
                   'e1071',
                   'mlbench',
                   'ggthemes',
                   'assertthat',
                   'kernlab',
                   'rpart',
                   'rattle',
                   'kknn',
                   'randomForest',
                   'nnet',
                   'RSNNS',
                   'Metrics',
                   'AUC',
                   'plotly',
                   'rBayesianOptimization')
)

---

B. Jupyter Notebooks

In case you want to run our amazing notebooks offline.

1. Install the latest release of Python

• https://www.python.org/ftp/python/3.6.5/python-3.6.5-amd64.exe (Windows)
• https://www.python.org/ftp/python/3.6.5/python-3.6.5-macosx10.9.pkg (Mac OS X 10.9 and later)

Do not forget to set the following environment variables on your device

• PATH must point to the directory of python.exe (e.g. C:\User\FolderWhereYouInstalledPython)
• PATH must point to the directory of python.exe (e.g. C:\User\FolderWhereYouInstalledPython\scripts)

And, in case you are working on a corporate device:

• set HTTP_PROXY to http://YOUR_ID:YOUR_UNIDIR_PWD@proxymil.internal.unicredit.eu:3128/
• and HTTPS_PROXY to https://YOUR_ID:YOUR_UNIDIR_PWD@proxymil.internal.unicredit.eu:3128/

2. Install visual Studio Build Tools

• https://www.visualstudio.com/it/thank-you-downloading-visual-studio/?sku=BuildTools&rel=15

3. From the CMD line (opened as administrator)

• If you wish to use jupyter notebooks

pip install jupyter

• If you prefer jupyterlab instead

pip install jupyterlab

4. Configure stuff in R

Do not forget to set the following environment variables on your device

• PATH must point to the directory of R.exe (e.g. C:\Program Files\R\R-3.4.1\bin)

install.packages(c('repr',
                   'IRdisplay',
                   'evaluate',
                   'crayon',
                   'pbdZMQ',
                   'devtools',
                   'uuid',
                   'digest'))

• Once you have installed the required packages, run this:

devtools::install_github('IRkernel/IRkernel')
IRkernel::installspec(user = FALSE)

• Only if the command above did not work, please download and unzip this and run the following:

devtools::install('C:\whatheverFolderYouHaveUnzippedTheFile\IRKernel')
IRkernel::installspec(user = FALSE)

5. From the CMD line (again, opened as administrator)

• Be sure to be in the right hard drive

C:

• Download the source files and unzip.

• Start jupyter/jupyterlab (a browser window should magically pop up). From there, you can source the .ipynb in the folder where you have downloaded it

jupyter-notebook
jupyter-lab

• To end the jupyter session (once you are done), hit CTRL+C twice in the CMD line

---

C. Neural Networks

Requires Python installation (see B.)

1. 64-bit OS

Sorry, this is a must have.

2. Interface R - Python

install.packages('reticulate')

3. Neural Nets

install.packages(c('tensorflow',
                   'keras')
)

tensorflow::install_tensorflow()
keras::install_keras() # this may throw an error but work anyway

4. Save and Load trained models

Run the follwing in the command line.

pip install h5py

---

D. Cloud

In case you do not manage to execute A. and B. you can run everything from https://rstudio.cloud