When I work as a data consultant, I often use Pandas in Jupyter to visualize the values taken from the DB to the text, so I will summarize the visualization around the final stage.
If you use the ones listed here, I think most of the work will be done. Refer to the Tutorial for how to use each. It's not that difficult.
matplotlib
Basic library. Everyone is using
seaborn
A library based on matplotlib that visualizes statistical analysis results in a visual manner. You can also write a dendrogram. Can be linked with pandas.
bokhe
A visualization library written in JS with full scratch. If you use matplotlib as it is, an image will be generated on Jupyter, but if you use bokhe, you can perform interactive operations such as scaling. It can be used quite conveniently in customer demos. I can make animations, but it's dull.
ggplot
Python version of R's ggplot2. Useful for visualizing the results of pandas DataFrame and stats model. I haven't gotten used to it yet, so for the time being, I feel like I should use seaborn.
plotly
A cloud-based visualization tool that can be used from python via API. There is a collaboration between Pandas and Jupyter. You can publish graphs on the Web. It's extremely sophisticated, and even 3D plots move slimy. If you just use it on Jupyter without publishing it, there seems to be no limit, so I would like to use it in the future. (https://plot.ly/python/offline/)
MLPD3
The interface of matplotlib that visualizes with D3.js as it is. It is convenient when you want to put the result of kneading with Jupyter as it is on the dash boat web application.
Recommended Posts