7 Python Data Visualization Libraries in 15 minutes

Key Takeaways

so you're working with some data in python and you want to make some visualizations well i've got news for you there's a lot of data visualization packages out there i mean a lot it can be a little overwhelming but no need to worry i'm gonna step through every single package known to man okay just the main packages used for plotting in python and by the end of this video you should have idea of what each of them do and how you can use them all right let's start checking out packages that doesn't sound right and here we are in our data vis kaggle notebook up first is the og of plotting in python yeah you know it matplotlib this package was initially released in 2003. were you born in 2003 because it's so old a lot of other things are built on it and there's a ton you can do first the imports import matplotlib import pi plot as plt this is just how you do it let's say we want to plot a line plot looks pretty nice kind of boring but nice wanna up the stakes why not your standard scatter plot just a few lines of code and we have colored dots ooh run border if you say rob i hate circles i got the solution for you matplotlib has tons of functionality you can change the points of the plot you could do a ton of stuff but rob i want to do some statistics well built into matplotlib you can make a histogram super easy look at these aardvark lengths oh wait no it's fake data [Music] but this is just the surface of what matplotlib can do it's basically a pen and paper for plotting you can make almost anything check out this awesome heat mat mesh grid [Music] not impressed with matplotlib i created this function that creates an entire football field if you're not impressed i don't know what will impress you so to summarize matplotlib one of the oldest libraries in python customization out the wazoo tons of other packages are built off of it but the downsides are it's hard to learn at first complex plots are going to be code heavy and even though it has the functionality it's not really meant for interactive plots next up we have seaborne little known fact did you know that seabourn was named after the character from west wing samuel norman seabourn that's why we import it as sns seabourn is great if you want to make some plots out of the box for statistical analysis stats major and what's the distribution you're just p hacking here we can use the seaborne rel plot function to show the difference between two groups in our data set wow not as much code as in matplotlib switch the style type to line and we have a beautiful line plot are those confidence intervals i see one of the great things about seaborn is you can look at the difference between different categories in your data without having to actually change the data itself time saver check out this swarm plot each stop represents an observation in our group and it split between different days of the week nifty and what about a violin plot well a violin plot's a great way to see the distribution compared between groups and even split within a group interested in just a normal old bar plot we got you covered seabourn can make those too with confidence intervals want to go beyond just a normal boring scatter plot seaborn can do a joint plot you can see distributions on the x and y axis of each different grouping and them colored in between that's pretty cool and one of my personal favorites is the pair plot have a bunch of numeric features that you want to compare you have kde plots on the diagonals scatter plots everywhere else what more could you ask for oh coloring by the different groups it's got that too and it may not seem like a big deal but in seabourn one of the features i love is that you can use the hue variable to add color to your plots so to summarize seaborne standard plots for statistical analysis easy to use out of the box it is built on matplotlib so it's really just a fancy matte plot lib there's much less ability to customize next up [Music] bouquet or boca is it bokeh or boca tell me what you think in the comments below i'm just gonna call it bouquet when we import bokeh we can set it up to work in a notebook just like that super simple bouquet specializes in interactive plots your standard line plot here but wait i can select things i can move it around this is amazing one line not enough for you how about multiple lines very cool okay let's pick it up a notch with bouquet layouts we could put multiple plots side by side but it goes even deeper we can plot two things at the same time linked look with this code if i select here it also selects it on the other side a lot of cool stuff you can do with these linked plots with a little bit more work you can make some pretty amazing plots in this plot we've added a bunch of circles but the size of the circles and the color are based on features in our data but wait there's more interactive plots are just the start with bokeh you can actually interact with the data live by adding things like a range slider check out this code a bunch of dots but we can filter based on this slider want to change the circle size on the fly why not the world's your oyster and of course the standard bar plot is always there if you need it we create a figure and use the bar to make something beautiful um looking at this plot makes me hungry once you become a pro with bouquet you can write really long plots like this that show unemployment numbers throughout the years i'm not gonna lie this heat map is pretty beautiful and you can even use bokeh to display those awesome network x graphs that you've been creating whoa when you hover over a circle you can actually see what it's linked to awesome so to summarize bouquet interactive plots almost infinite ways you can customize bokeh plots for me bokeh reminds me of the map plot lib equivalent but for interactive plots the downside though is it's like matplotlib so it requires a lot of code and the time has come we're gonna talk about plotly express i know you've been holding your breath so just let it go now you can breathe i'm really actually surprised plotly express only has 657 github stars go and give it a star importing plotly express is as simple as import as px plotley express like bokeh allows you to create interactive plots but the difference is they're a lot easier to create using plotly express scatter we can create a scatter plot with just a few lines of code oh look at these petal widths but i want a grid of plots you say well look no further just provide a facet column and facet row variable and you've got yourself a grid of scatter plots what about line plots we got you covered uh check out australia and new zealand oh australia is winning all this in just one line of code sign me up and it wouldn't be complete without some bar plots just like bokeh it's fully interactive so we can turn things off if we don't like it oh but i can't turn them both off because i kind of don't want to see this plot at all remember seaborne's pair plot well plotly express has the scatter plot matrix not a lot of code here and we see the comparison so this has the upside of being interactive but i tend to prefer seaborn for this type of plot there are also a handful of custom plots that are just gorgeous check out this parallel coordinates plot and one of my personal favorites the parallel categories plot huh different groupings and how they link oh you can hover over it amazing i'm sold i'm sold i'm sold i'm sold i'm sold here's a perfect example of using plotly express to make a scatter plot colors sizes all of them depending on the data whoa let's only look at oceania australia and new zealand again and for those of you who prefer pie charts the sunburst plot is sort of like a pie chart kind of just don't make pie charts just like seabourn again plotly express can do different comparison between categories check out these violin plots not too much code here to create it too and something you might not always use but if you ever have a data set with latitude and longitude the scatter map box function can make some beautiful maps two dimensions not enough for you well we can go 3d with the scattered 3d plot with plotly express i feel like i'm in the matrix so to summarize plotly express two thumbs up for interactive plots it's kind of like seaborne but interactive and those maps i mean i don't know what else to say if you're not impressed by that map just close the video right now just close it i don't want you here also similar to seabourn it's going to be hard to customize your plots there is the plotly back end and dash but that's for another video the next two libraries are slightly lesser known so let's talk about plot nine if you've used r before you probably know gg plot two it can make some gorgeous charts so plot nine is basically gg plot and python the way you write the plots is different than anything else that i've seen but it's very similar to what you would do in r so if you're familiar with r this might be the package for you we got scatter plots with regression lines and standard error can't be beat one thing i gotta say is the default colors in plot9 and ggplot are really nice with these few lines of code you can make this bar plot with some lines hey check out want to do some analysis of greek letters you can make a pretty bar plot just a few lines of code i like it so to summarize plot nine basically gg plot two positives the plots are pretty the syntax is a little unique but if you're used to r you might like it downside is if you're not used to it it takes a little bit of work to understand and i gotta say the docks aren't that great so if you're looking to do anything custom you might be up creek without a paddle let's talk about alteir to be honest i haven't used altair that much and i don't really understand why it's different but it is different and maybe if you understand it you can explain it to me so the positives are the positives are that it has declarative visualization and the negatives are that it has declarative visualization and last up we have pandas you didn't think i would get through a video without talking about pandas did you that's right pandas can do plots just take your data frame use the plot method and you can make it make a cool visualization like this in milliseconds but rob you say this is just matplotlib i know but it saves you time because it's already in pandas show me a scatter plot there it is pandas can even do box plots these leave something to be desired area plots are actually pretty cool and i use these a lot it's basically a line plot filled in underneath so there you have it as quickly and hopefully entertainingly as i could i showed you every python package commonly used for visualization if i missed one please let me know in the comments below we talked about matplotlib seaborn bokeh plotly express plot altair and pandas everything you would ever ask for for plotting with python my name is rob i'm a data scientist and kaggle grandmaster if you enjoyed this video please pretty please subscribe and like the video too oh also you can find me on twitch where i stream time to time okay that's it see ya bye using plotly expl using pr plotly exploit using plot leaks