Contents

This summarizes my Python setup for data science/finance/economics research. It includes some of the useful Python packages, JupyterLab extensions, and programs I find useful in my development.

It assumes the following:

Basic knowledge of using conda for installing Python packages.
Ensure that conda-forge is one of the default installation channels for Python packages.

Command line multiplexer

Command line multiplxer (tmux!)

I often have a screen multiplexer for my console screen (i.e., command line prompt). This is because I might be updating this website, doing some coding with Jupyter lab, or need to install some Python packages etc.

Windows

For multiplexer support for Anaconda Prompt, install ConEmu.
Under Settings -> Startup -> Command line.
Key in the file location of where the Anaconda Prompt batch file is (e.g., "%windir%\System32\cmd.exe "/K" C:\Users\<YOUR USERNAME>\Anaconda3\Scripts\activate.bat").

Tip

If uncertain of the above entry, right-click your Anaconda Prompt icon and Open File Location.
Right-click the Anaconda Prompt shortcut.
Go Shortcut -> Target.
In Target, you will find the correct entry that is required such that when ConEmu starts up a new instance of the Console, it will automatically open up the Anaconda Prompt instead of the Windows Command Prompt.

Linux

I use tmux with my handy tmux cheatsheet. For more instructions on how to install it in Linux, see Setting up the Pixelbook/Chromebook for Python programming.

Installing Python packages

Use the following command in Terminal to install packages.

$ conda install <PKG1> <PKG2> <PKG3>

Core data analytics

Package Name	Description
numpy	Matrix operations
scipy	Scientific operations
pandas	DataFrame operations
statsmodels	Statistical models
matplotlib	Data visualization

Finance

Package Name	Description
pandas_datareader	Reading financial data from the web
arch	Volatility modelling
cvxopt	Core optimization
cvxpy	Wrapper to "nicely" interface into cvxopt
prophet	Facebook's timeseries forecasting tool

Data visualization

Package Name	Description
seaborn	Colorful data visualization
bokeh	Interactive data visualization
geoplotlib	Create maps and geographical visualization
Gleam	Similar to R's Shiny web App
Altair	New framework for plotting graphics
plotnine	Similar to R's ggplot2 using grammer of graphics

Machine & deep learning

Package Name	Description
scikit-learn	Core machine learning package
xgboost	The first publicly available gradient boosting package. Released by Tianqi Chen (University of Washington, Seattle)
lightgbm	Gradient Boosted Decision Trees package (Microsoft)
catboost	Gradient Boosting Decision Trees package (Yandex)
keras	High-level neural networks API
Tensorflow	Deep learning package from Google

Natural language processing (NLP)

Package Name	Description
nltk	General NLP tasks
textblog	Creating NLP prototypes quickly
gensim	NLP applications for topic modelling, document similarity, etc.
scrapy	Web scraping
spacy	Production-level NLP library

Favourite JupyterLab extensions

See JupyterLab extensions for more details and GitHub for a full list of available extensions for JupyterLab.

Install nodejs

$ conda install -c conda-forge nodejs

Installing Jupyterlab extensions in Linux as follows:

$ jupyter labextension install @jupyterlab/<EXT_NAME>
$ jupyter labextension install jupyterlab_nbmetadata

You can install plugins in Jupyter lab by clicking the jigsaw icon on the menu bar on the menu bar on the right of the editor.
The following are my favourite Jupyter lab plugins.

Extension name	Description
toc	Table of contents
jupyterlab_variableInspector	Variable inspector in Jupyter Lab
jupyterlab_nbmetadata	Allows you to edit the notebook metadata
jupyterlab_go_to_definition	Jump to definition of a variable or function in JupyterLab notebook and file editor

Installing Jupyter notebook extensions

$ conda install -c conda-forge jupyter_contrib_nbextensions

Additional Jupyter commands

jupyter notebook list: Lists down all running Jupyter notebooks and which portIDs they are running on (e.g., 8889, 8888)
jupyter notebook stop <portID>: Stops the Jupyter notebook from running once the portID is provided.

Comments