A quick analysis of Steemit tags data using Python

My first real post! You can learn more about me here

Whilst exploring steemit and learning more about how everything works, I tried looking at things the way I enjoy most, through data!

This a quick analysis I did of the tags summary data using Python (would love to learn if there are better ways of formatting code in posts!). I tried getting a feel for post, comments and payouts on steemit based the post tags.

I've left the code inline if you're interested!

Here is a link to the full Jupyter notebook as well (including the raw data) https://gist.github.com/michael-erasmus/4e7334537a445f3e5830cde9d7db644c

If you enjoyed this post and would like to see more stuff like this, let me know! (and upvote!)

%matplotlib inline
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.ticker import FuncFormatter
import seaborn as sns

sns.set_style('whitegrid')

---

Read in data derived from trending topics list

This was collected on 2017-07-21 22:46:42 (UTC)

df = pd.read_csv('tags.csv')
df.head()

	Tag	Posts	Comments	Payouts
0	aceh	4399	805	55180.578
1	adventure	1474	494	147060.430
2	advice	554	210	21937.755
3	altcoin	627	105	19558.774
4	amazing	617	85	2204.555

---

A quick look at the summary stats

df.describe()

	Posts	Comments	Payouts
count	249.000000	249.000000	2.490000e+02
mean	1931.751004	1866.923695	1.743020e+05
std	4419.145897	5418.708141	4.639500e+05
min	1.000000	32.000000	0.000000e+00
25%	422.000000	206.000000	1.679843e+04
50%	667.000000	408.000000	4.425472e+04
75%	1387.000000	1074.000000	1.194865e+05
max	42928.000000	43884.000000	4.091146e+06

---

Let's see a quick distribution of Posts, Comments and Payouts

---

fig, (ax1, ax2, ax3) = plt.subplots(ncols=3, figsize=(30,10))
sns.distplot(df.Posts, ax=ax1, color='b')
sns.distplot(df.Comments, ax=ax2, color='r')
sns.distplot(df.Payouts, color='g')

---

Let's look at the top 20 tags in terms of Posts, Comments, and Payouts

top = lambda n, by: df.sort_values(by=by, ascending=False).head(n)
payout_formatter = FuncFormatter(lambda x,p: '%1.1fM' % (x*1e-6))

fig, (ax1,ax2,ax3) = plt.subplots(figsize=(20,15), ncols=3)

sns.barplot(y='Tag', x='Posts',
            palette=sns.light_palette("blue", n_colors=20, reverse=True),
            data=top(20, "Posts"), 
            ax=ax1)
ax1.set(title="Posts", xlabel='')

sns.barplot(y='Tag', x='Comments',
            palette=sns.light_palette("red", n_colors=20, reverse=True),
            data=top(20, "Comments"), 
            ax=ax2)
ax2.set(title="Comments", xlabel='', ylabel='')

sns.barplot(y='Tag', x='Payouts',
            palette=sns.light_palette("green", n_colors=20, reverse=True),
            data=top(20, "Payouts"), 
            ax=ax3)

ax3.set(title="Payouts in SBD", xlabel='', ylabel='')
ax3.xaxis.set_major_formatter(payout_formatter)

fig.suptitle('Top 20 Tags in terms of Posts, Comments and Payouts')

---

Let's look at the Payout's in terms of posts and comments

df = df.sort_values(by="Payouts", ascending=False)

fig, (ax1,ax2) = plt.subplots(figsize=(20,10), ncols=2, sharey=True)

sns.regplot(x='Posts', 
            y='Payouts', 
            scatter_kws={'alpha':0.5}, 
            data=df, 
            ax=ax1)

for i, tag in enumerate(df.Tag.head(20)):
    ax1.annotate(tag,(list(df.Posts)[i],list(df.Payouts)[i]))
    
ax1.yaxis.set_major_formatter(payout_formatter)


sns.regplot(x='Comments', 
            y='Payouts', 
            scatter_kws={'alpha':0.5}, 
            data=df, 
            ax=ax2)

#Annotate the top 20 in terms of payout
for i, tag in enumerate(df.Tag.head(20)):
    ax2.annotate(tag,(list(df.Comments)[i],list(df.Payouts)[i]))
    
ax2.yaxis.set_major_formatter(payout_formatter)
ax2.set(ylabel='')

---

What's the payout per post/comment for tags?

It's important to note that this is not the same as the "Average Post/Comment Payout"!

Because we only have the Total Payout without knowing how much is allocated for Posts and Comments, the exact ratio might not be too useful, but can still give us a rough idea how the tags compare against each other.

df['payout_per_post'] = df.Payouts / df.Posts

df[['Tag', 'Posts', 'payout_per_post']] \
    .sort_values(by='payout_per_post', ascending=False)\
    .head(20) \
    .reset_index() \
    .drop('index', axis=1)

	Tag	Posts	payout_per_post
0	bitshares	286	1105.515846
1	steemfest	125	925.923792
2	witness-category	113	844.448761
3	meetup	112	643.752375
4	beyondbitcoin	994	500.433907
5	steem-help	211	469.506502
6	steem-pocalypse	38	411.794526
7	curation	371	368.283876
8	security	530	353.183917
9	coinkorea	481	339.354857
10	crypto-news	984	326.677209
11	gridcoin	75	319.262480
12	charlesfuchs	8	317.167500
13	introduceyourself	4430	306.785769
14	stats	402	296.285604
15	marketing	560	274.113300
16	economics	470	264.304479
17	steem	10607	253.045986
18	steemvoter	114	252.183789
19	eos	362	251.570713

Let's do the same thing for comments

df['payout_per_comment'] = df.Payouts / df.Comments

df[['Tag', 'Comments', 'payout_per_comment']] \
    .sort_values(by='payout_per_comment', ascending=False) \
    .head(20) \
    .reset_index() \
    .drop('index', axis=1)

	Tag	Comments	payout_per_comment
0	creativity	163	1481.952816
1	minnowsunite	170	1479.143265
2	future	35	1348.421343
3	finance	48	1218.871104
4	marketing	142	1081.010197
5	review	66	1028.849197
6	security	215	870.639423
7	work	87	633.268655
8	stats	190	626.877963
9	creative	143	610.995790
10	beyondbitcoin	816	609.597186
11	china	49	605.306388
12	recipe	161	587.820820
13	government	104	578.813019
14	anarchism	295	568.976651
15	girls	127	507.826693
16	fun	246	470.909626
17	recipes	256	466.744328
18	painting	237	466.703169
19	tutorial	132	452.375341