StatsExchange - Predicting users' reputations¶

In this notebook, we will use relational learning techniques to predict users' reputation on StackExchange

Summary:

Prediction type: Regression model
Domain: Internet
Prediction target: Reputation
Population size: 41793

Background¶

StatsExchange is a website similar to StackOverflow, but based on statistics and machine learning. Much like StackOverflow, it has a complicated system of calculating users' reputation. In this notebook, we will benchmark relational learning techniques based on there ability to reverse-engineer said system.

The dataset has been downloaded from the CTU Prague relational learning repository (Motl and Schulte, 2015)(Now residing at relational-data.org.).

We will benchmark getML 's feature learning algorithms against featuretools, an open-source implementation of the propositionalization algorithm, similar to getML's FastProp.

Analysis¶

Let's get started with the analysis and set up your session:

In [1]:

  Copied!     
 
%pip install -q "getml==1.5.0" "featuretools==1.31.0" "ipywidgets==8.1.5"
%pip install -q "getml==1.5.0" "featuretools==1.31.0" "ipywidgets==8.1.5"

In [1]:

  Copied!     
 
import os
import warnings

import pandas as pd

import featuretools
import woodwork as ww
import getml

os.environ["PYARROW_IGNORE_TIMEZONE"] = "1"
warnings.simplefilter(action='ignore', category=FutureWarning)

print(f"getML API version: {getml.__version__}\n")
import os import warnings import pandas as pd import featuretools import woodwork as ww import getml os.environ["PYARROW_IGNORE_TIMEZONE"] = "1" warnings.simplefilter(action='ignore', category=FutureWarning) print(f"getML API version: {getml.__version__}\n")

getML API version: 1.5.0

In [2]:

  Copied!     
 
getml.engine.launch(allow_remote_ips=True, token='token')
getml.engine.set_project('stats')
getml.engine.launch(allow_remote_ips=True, token='token') getml.engine.set_project('stats')

Launching ./getML --allow-push-notifications=true --allow-remote-ips=true --home-directory=/home/user --in-memory=true --install=false --launch-browser=true --log=false --token=token in /home/user/.getML/getml-1.5.0-x64-linux...
Launched the getML Engine. The log output will be stored in /home/user/.getML/logs/20240912151014.log.
  Loading pipelines... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00

Connected to project 'stats'.

1. Loading data¶

1.1 Download from source¶

We begin by downloading the data:

In [4]:

  Copied!     
 
conn = getml.database.connect_mysql(
    host="db.relational-data.org",
    dbname="stats",
    port=3306,
    user="guest",
    password="relational"
)

conn
conn = getml.database.connect_mysql( host="db.relational-data.org", dbname="stats", port=3306, user="guest", password="relational" ) conn

Out[4]:

Connection(dbname='stats', dialect='mysql', host='db.relational-data.org', port=3306)

In [5]:

  Copied!     
 
def load_if_needed(name):
    """
    Loads the data from the relational learning
    repository, if the data frame has not already
    been loaded.
    """
    if not getml.data.exists(name):
        data_frame = getml.data.DataFrame.from_db(
            name=name,
            table_name=name,
            conn=conn
        )
        data_frame.save()
    else:
        data_frame = getml.data.load_data_frame(name)
    return data_frame
def load_if_needed(name): """ Loads the data from the relational learning repository, if the data frame has not already been loaded. """ if not getml.data.exists(name): data_frame = getml.data.DataFrame.from_db( name=name, table_name=name, conn=conn ) data_frame.save() else: data_frame = getml.data.load_data_frame(name) return data_frame

In [6]:

  Copied!     
 
badges = load_if_needed("badges")
posts = load_if_needed("posts")
users = load_if_needed("users")
votes = load_if_needed("votes")
badges = load_if_needed("badges") posts = load_if_needed("posts") users = load_if_needed("users") votes = load_if_needed("votes")

In [7]:

  Copied!     
 
badges
badges

Out[7]:

name	Id	UserId	Name	Date
role	unused_float	unused_float	unused_string	unused_string
0	1	5	Teacher	2010-07-19 19:39:07
1	2	6	Teacher	2010-07-19 19:39:07
2	3	8	Teacher	2010-07-19 19:39:07
3	4	23	Teacher	2010-07-19 19:39:07
4	5	36	Teacher	2010-07-19 19:39:07
	...	...	...	...
79846	92236	55744	Student	2014-09-13 23:25:21
79847	92237	1118	Nice Answer	2014-09-14 00:09:35
79848	92238	1118	Enlightened	2014-09-14 01:18:29
79849	92239	55746	Student	2014-09-14 01:41:18
79850	92240	12597	Autobiographer	2014-09-14 02:31:28

79851 rows x 4 columns
memory usage: 4.96 MB
name: badges
type: getml.DataFrame

In [8]:

  Copied!     
 
posts
posts

Out[8]:

name	Id	PostTypeId	AcceptedAnswerId	Score	ViewCount	OwnerUserId	AnswerCount	CommentCount	FavoriteCount	LastEditorUserId	ParentId	CreaionDate	Body	LasActivityDate	Title	Tags	LastEditDate	CommunityOwnedDate	ClosedDate	OwnerDisplayName	LastEditorDisplayName
role	unused_float	unused_float	unused_float	unused_float	unused_float	unused_float	unused_float	unused_float	unused_float	unused_float	unused_float	unused_string	unused_string	unused_string	unused_string	unused_string	unused_string	unused_string	unused_string	unused_string	unused_string
0	1	1	15	23	1278	8	5	1	14	nan	nan	2010-07-19 19:12:12	<p>How should I elicit prior dis...	2010-09-15 21:08:26	Eliciting priors from experts	<bayesian><prior><elicitation>	NULL	NULL	NULL	NULL	NULL
1	2	1	59	22	8198	24	7	1	8	88	nan	2010-07-19 19:12:57	<p>In many different statistical...	2012-11-12 09:21:54	What is normality?	<distributions><normality>	2010-08-07 17:56:44	NULL	NULL	NULL	NULL
2	3	1	5	54	3613	18	19	4	36	183	nan	2010-07-19 19:13:28	<p>What are some valuable Statis...	2013-05-27 14:48:36	What are some valuable Statistic...	<software><open-source>	2011-02-12 05:50:03	2010-07-19 19:13:28	NULL	NULL	NULL
3	4	1	135	13	5224	23	5	2	2	nan	nan	2010-07-19 19:13:31	<p>I have two groups of data. E...	2010-09-08 03:00:19	Assessing the significance of di...	<distributions><statistical-sign...	NULL	NULL	NULL	NULL	NULL
4	5	2	nan	81	nan	23	nan	3	nan	23	3	2010-07-19 19:14:43	<p>The R-project</p> <p><a href...	2010-07-19 19:21:15	NULL	NULL	2010-07-19 19:21:15	2010-07-19 19:14:43	NULL	NULL	NULL
	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
91971	115374	2	nan	2	nan	805	nan	2	nan	805	115367	2014-09-13 23:45:39	<p>This grew too long for a comm...	2014-09-14 02:05:41	NULL	NULL	2014-09-14 02:05:41	NULL	NULL	NULL	NULL
91972	115375	1	nan	0	9	49365	1	0	nan	nan	nan	2014-09-13 23:46:05	<p>Assume a classification probl...	2014-09-14 02:09:23	Detecting a consistent pattern i...	<classification><cross-validatio...	NULL	NULL	NULL	NULL	NULL
91973	115376	1	nan	1	5	55746	0	2	nan	7290	nan	2014-09-14 01:27:54	<p>My goal is to create a formul...	2014-09-14 01:40:55	How to project video viewcount b...	<summary-statistics><median><evi...	2014-09-14 01:40:55	NULL	NULL	NULL	NULL
91974	115377	2	nan	0	nan	805	nan	0	nan	805	115358	2014-09-14 02:03:28	<p>As a practical answer to the ...	2014-09-14 02:54:13	NULL	NULL	2014-09-14 02:54:13	NULL	NULL	NULL	NULL
91975	115378	2	nan	0	nan	7250	nan	0	nan	nan	115375	2014-09-14 02:09:23	<p>Decision trees are notoriousl...	2014-09-14 02:09:23	NULL	NULL	NULL	NULL	NULL	NULL	NULL

91976 rows x 21 columns
memory usage: 128.41 MB
name: posts
type: getml.DataFrame

In [9]:

  Copied!     
 
users
users

Out[9]:

name	Id	Reputation	Views	UpVotes	DownVotes	AccountId	Age	CreationDate	DisplayName	LastAccessDate	WebsiteUrl	Location	AboutMe	ProfileImageUrl
role	unused_float	unused_float	unused_float	unused_float	unused_float	unused_float	unused_float	unused_string	unused_string	unused_string	unused_string	unused_string	unused_string	unused_string
0	-1	1	0	5007	1920	-1	nan	2010-07-19 06:55:26	Community	2010-07-19 06:55:26	http://meta.stackexchange.com/	on the server farm	<p>Hi, I'm not really a person.<...	NULL
1	2	101	25	3	0	2	37	2010-07-19 14:01:36	Geoff Dalgas	2013-11-12 22:07:23	http://stackoverflow.com	Corvallis, OR	<p>Developer on the StackOverflo...	NULL
2	3	101	22	19	0	3	35	2010-07-19 15:34:50	Jarrod Dixon	2014-08-08 06:42:58	http://stackoverflow.com	New York, NY	<p><a href="http://blog.stackove...	NULL
3	4	101	11	0	0	1998	28	2010-07-19 19:03:27	Emmett	2014-01-02 09:31:02	http://minesweeperonline.com	San Francisco, CA	<p>currently at a startup in SF<...	http://i.stack.imgur.com/d1oHX.j...
4	5	6792	1145	662	5	54503	35	2010-07-19 19:03:57	Shane	2014-08-13 00:23:47	http://www.statalgo.com	New York, NY	<p>Quantitative researcher focus...	NULL
	...	...	...	...	...	...	...	...	...	...	...	...	...	...
40320	55743	1	0	0	0	5026902	nan	2014-09-13 21:03:50	AussieMeg	2014-09-13 21:18:52	NULL	NULL	NULL	http://graph.facebook.com/665821...
40321	55744	6	1	0	0	5026998	nan	2014-09-13 21:39:30	Mia Maria	2014-09-13 21:39:30	NULL	NULL	NULL	NULL
40322	55745	101	0	0	0	481766	nan	2014-09-13 23:45:27	tronbabylove	2014-09-13 23:45:27	NULL	United States	NULL	https://www.gravatar.com/avatar/...
40323	55746	106	1	0	0	976289	nan	2014-09-14 00:29:41	GPP	2014-09-14 02:05:17	NULL	NULL	<p>Stats noobie, product, market...	https://www.gravatar.com/avatar/...
40324	55747	1	0	0	0	5027354	nan	2014-09-14 01:01:44	Shivam Agrawal	2014-09-14 01:19:04	NULL	India	<p>Maths Enthusiast </p>	https://lh4.googleusercontent.co...

40325 rows x 14 columns
memory usage: 10.32 MB
name: users
type: getml.DataFrame

In [10]:

  Copied!     
 
votes
votes

Out[10]:

name	Id	PostId	VoteTypeId	UserId	BountyAmount	CreationDate
role	unused_float	unused_float	unused_float	unused_float	unused_float	unused_string
0	1	3	2	nan	nan	2010-07-19
1	2	2	2	nan	nan	2010-07-19
2	3	5	2	nan	nan	2010-07-19
3	4	5	2	nan	nan	2010-07-19
4	5	3	2	nan	nan	2010-07-19
	...	...	...	...	...	...
328059	386254	26088	2	nan	nan	2014-09-14
328060	386255	26088	5	31466	nan	2014-09-14
328061	386256	115374	2	nan	nan	2014-09-14
328062	386257	115368	2	nan	nan	2014-09-14
328063	386258	115369	2	nan	nan	2014-09-14

328064 rows x 6 columns
memory usage: 19.36 MB
name: votes
type: getml.DataFrame

1.2 Prepare data for getML¶

getML requires that we define roles for each of the columns.

In [11]:

  Copied!     
 
badges.set_role(["Id", "UserId"], getml.data.roles.join_key)
badges.set_role("Date", getml.data.roles.time_stamp)
badges.set_role("Name", getml.data.roles.categorical)

badges
badges.set_role(["Id", "UserId"], getml.data.roles.join_key) badges.set_role("Date", getml.data.roles.time_stamp) badges.set_role("Name", getml.data.roles.categorical) badges

Out[11]:

name	Date	Id	UserId	Name
role	time_stamp	join_key	join_key	categorical
unit	time stamp, comparison only
0	2010-07-19 19:39:07	1	5	Teacher
1	2010-07-19 19:39:07	2	6	Teacher
2	2010-07-19 19:39:07	3	8	Teacher
3	2010-07-19 19:39:07	4	23	Teacher
4	2010-07-19 19:39:07	5	36	Teacher
	...	...	...	...
79846	2014-09-13 23:25:21	92236	55744	Student
79847	2014-09-14 00:09:35	92237	1118	Nice Answer
79848	2014-09-14 01:18:29	92238	1118	Enlightened
79849	2014-09-14 01:41:18	92239	55746	Student
79850	2014-09-14 02:31:28	92240	12597	Autobiographer

79851 rows x 4 columns
memory usage: 1.60 MB
name: badges
type: getml.DataFrame

In [12]:

  Copied!     
 
posts.set_role(["Id", "AcceptedAnswerId", "OwnerUserId"], getml.data.roles.join_key)
posts.set_role(["Score", "ViewCount", "AnswerCount", "CommentCount", "FavoriteCount", "LastEditorUserId"], getml.data.roles.numerical)
posts.set_role(["PostTypeId"], getml.data.roles.categorical)

posts.drop(posts.roles.unused)
posts.set_role(["Id", "AcceptedAnswerId", "OwnerUserId"], getml.data.roles.join_key) posts.set_role(["Score", "ViewCount", "AnswerCount", "CommentCount", "FavoriteCount", "LastEditorUserId"], getml.data.roles.numerical) posts.set_role(["PostTypeId"], getml.data.roles.categorical) posts.drop(posts.roles.unused)

Out[12]:

name	Id	AcceptedAnswerId	OwnerUserId	PostTypeId	Score	ViewCount	AnswerCount	CommentCount	FavoriteCount	LastEditorUserId
role	join_key	join_key	join_key	categorical	numerical	numerical	numerical	numerical	numerical	numerical
0	1	15	8	1	23	1278	5	1	14	nan
1	2	59	24	1	22	8198	7	1	8	88
2	3	5	18	1	54	3613	19	4	36	183
3	4	135	23	1	13	5224	5	2	2	nan
4	5	nan	23	2	81	nan	nan	3	nan	23
...	...	...	...	...	...	...	...	...	...	...

91976 rows
type: getml.data.View

In [13]:

  Copied!     
 
users.set_role("Reputation", getml.data.roles.target)
users.set_role(["CreationDate", "LastAccessDate"], getml.data.roles.time_stamp)
users.set_role(["Id"], getml.data.roles.join_key)
users.set_role(["Views", "UpVotes", "DownVotes"], getml.data.roles.numerical)

users.drop(users.roles.unused)
users.set_role("Reputation", getml.data.roles.target) users.set_role(["CreationDate", "LastAccessDate"], getml.data.roles.time_stamp) users.set_role(["Id"], getml.data.roles.join_key) users.set_role(["Views", "UpVotes", "DownVotes"], getml.data.roles.numerical) users.drop(users.roles.unused)

Out[13]:

name	CreationDate	LastAccessDate	Id	Reputation	Views	UpVotes	DownVotes
role	time_stamp	time_stamp	join_key	target	numerical	numerical	numerical
unit	time stamp, comparison only	time stamp, comparison only
0	2010-07-19 06:55:26	2010-07-19 06:55:26	-1	1	0	5007	1920
1	2010-07-19 14:01:36	2013-11-12 22:07:23	2	101	25	3	0
2	2010-07-19 15:34:50	2014-08-08 06:42:58	3	101	22	19	0
3	2010-07-19 19:03:27	2014-01-02 09:31:02	4	101	11	0	0
4	2010-07-19 19:03:57	2014-08-13 00:23:47	5	6792	1145	662	5
...	...	...	...	...	...	...	...

40325 rows
type: getml.data.View

In [14]:

  Copied!     
 
votes.set_role(["CreationDate"], getml.data.roles.time_stamp)
votes.set_role(["Id", "PostId"], getml.data.roles.join_key)
votes.set_role(["VoteTypeId"], getml.data.roles.categorical)
votes.set_role(["BountyAmount"], getml.data.roles.numerical)

votes.drop(users.roles.unused)
votes.set_role(["CreationDate"], getml.data.roles.time_stamp) votes.set_role(["Id", "PostId"], getml.data.roles.join_key) votes.set_role(["VoteTypeId"], getml.data.roles.categorical) votes.set_role(["BountyAmount"], getml.data.roles.numerical) votes.drop(users.roles.unused)

Out[14]:

name	CreationDate	Id	PostId	VoteTypeId	BountyAmount	UserId
role	time_stamp	join_key	join_key	categorical	numerical	unused_float
unit	time stamp, comparison only
0	2010-07-19	1	3	2	nan	nan
1	2010-07-19	2	2	2	nan	nan
2	2010-07-19	3	5	2	nan	nan
3	2010-07-19	4	5	2	nan	nan
4	2010-07-19	5	3	2	nan	nan
...	...	...	...	...	...	...

328064 rows
type: getml.data.View

In [15]:

  Copied!     
 
split = getml.data.split.random(train=0.8, test=0.2)
split = getml.data.split.random(train=0.8, test=0.2)

In [16]:

  Copied!     
 
container = getml.data.Container(population=users, split=split)

container.add(badges=badges, posts=posts, votes=votes)

container
container = getml.data.Container(population=users, split=split) container.add(badges=badges, posts=posts, votes=votes) container

Out[16]:

population

	subset	name	rows	type
0	test	users	7992	View
1	train	users	32333	View

peripheral

	name	rows	type
0	badges	79851	DataFrame
1	posts	91976	DataFrame
2	votes	328064	DataFrame

2. Predictive modeling¶

We loaded the data and defined the roles and units. Next, we create a getML pipeline for relational learning.

2.1 Define relational model¶

In [17]:

  Copied!     
 
dm = getml.data.DataModel(population=users.to_placeholder())

dm.add(getml.data.to_placeholder(badges=badges, posts=[posts]*2, votes=votes))

dm.population.join(
    dm.badges,
    on=("Id", "UserId"),
)

dm.population.join(
    dm.posts[0],
    on=("Id", "OwnerUserId"),
)

dm.posts[0].join(
    dm.posts[1],
    on=("Id", "AcceptedAnswerId"),
    relationship=getml.data.relationship.one_to_one,
)

dm.posts[0].join(
    dm.votes,
    on=("Id", "PostId"),
)

dm
dm = getml.data.DataModel(population=users.to_placeholder()) dm.add(getml.data.to_placeholder(badges=badges, posts=[posts]*2, votes=votes)) dm.population.join( dm.badges, on=("Id", "UserId"), ) dm.population.join( dm.posts[0], on=("Id", "OwnerUserId"), ) dm.posts[0].join( dm.posts[1], on=("Id", "AcceptedAnswerId"), relationship=getml.data.relationship.one_to_one, ) dm.posts[0].join( dm.votes, on=("Id", "PostId"), ) dm

Out[17]:

diagram

staging

	data frames	staging table
0	users	USERS__STAGING_TABLE_1
1	badges	BADGES__STAGING_TABLE_2
2	posts, posts	POSTS__STAGING_TABLE_3
3	votes	VOTES__STAGING_TABLE_4

2.2 getML pipeline¶

Set-up the feature learner & predictor

We use the relboost algorithms for this problem. Because of the large number of keywords, we regularize the model a bit by requiring a minimum support for the keywords (min_num_samples).

In [18]:

  Copied!     
 
mapping = getml.preprocessors.Mapping()

fast_prop = getml.feature_learning.FastProp(
    aggregation=["AVG", "SUM"],
    loss_function=getml.feature_learning.loss_functions.SquareLoss,
    num_threads=1,
    n_most_frequent=5,
    
)

relboost = getml.feature_learning.Relboost(
    loss_function=getml.feature_learning.loss_functions.SquareLoss,
    num_threads=1,
    num_subfeatures=10,
)

predictor = getml.predictors.XGBoostRegressor(n_jobs=1)
mapping = getml.preprocessors.Mapping() fast_prop = getml.feature_learning.FastProp( aggregation=["AVG", "SUM"], loss_function=getml.feature_learning.loss_functions.SquareLoss, num_threads=1, n_most_frequent=5, ) relboost = getml.feature_learning.Relboost( loss_function=getml.feature_learning.loss_functions.SquareLoss, num_threads=1, num_subfeatures=10, ) predictor = getml.predictors.XGBoostRegressor(n_jobs=1)

Build the pipeline

In [19]:

  Copied!     
 
pipe1 = getml.Pipeline(
    tags=['fast_prop'],
    data_model=dm,
    feature_learners=[fast_prop],
    predictors=[predictor],
    include_categorical=True,
)

pipe1
pipe1 = getml.Pipeline( tags=['fast_prop'], data_model=dm, feature_learners=[fast_prop], predictors=[predictor], include_categorical=True, ) pipe1

Out[19]:

Pipeline(data_model='users',
         feature_learners=['FastProp'],
         feature_selectors=[],
         include_categorical=True,
         loss_function='SquareLoss',
         peripheral=['badges', 'posts', 'votes'],
         predictors=['XGBoostRegressor'],
         preprocessors=[],
         share_selected_features=0.5,
         tags=['fast_prop'])

In [20]:

  Copied!     
 
pipe2 = getml.Pipeline(
    tags=['relboost'],
    data_model=dm,
    preprocessors=[mapping],
    feature_learners=[relboost],
    predictors=[predictor],    
    include_categorical=True,
)

pipe2
pipe2 = getml.Pipeline( tags=['relboost'], data_model=dm, preprocessors=[mapping], feature_learners=[relboost], predictors=[predictor], include_categorical=True, ) pipe2

Out[20]:

Pipeline(data_model='users',
         feature_learners=['Relboost'],
         feature_selectors=[],
         include_categorical=True,
         loss_function='SquareLoss',
         peripheral=['badges', 'posts', 'votes'],
         predictors=['XGBoostRegressor'],
         preprocessors=['Mapping'],
         share_selected_features=0.5,
         tags=['relboost'])

2.3 Model training¶

In [21]:

  Copied!     
 
pipe1.check(container.train)
pipe1.check(container.train)

Checking data model...

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
⠋ Checking... ━━━━━━━━━━━━━━╸                           37% • 00:01

  Checking... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00

The pipeline check generated 3 issues labeled INFO and 3 issues labeled WARNING.

Out[21]:

	type	label	message
0	WARNING	COLUMN SHOULD BE UNUSED	All non-NULL entries in column 't3__posttypeid' in POSTS__STAGING_TABLE_3 are equal to each other. You should consider setting its role to unused_string or using it for comparison only (you can do the latter by setting a unit that contains 'comparison only').
1	WARNING	COLUMN SHOULD BE UNUSED	93.147125% of all entries in column 't3__favoritecount' in POSTS__STAGING_TABLE_3 are NULL values. You should consider setting its role to unused_float.
2	WARNING	COLUMN SHOULD BE UNUSED	99.468396% of all entries in column 'bountyamount' in VOTES__STAGING_TABLE_4 are NULL values. You should consider setting its role to unused_float.
3	INFO	FOREIGN KEYS NOT FOUND	When joining USERS__STAGING_TABLE_1 and BADGES__STAGING_TABLE_2 over 'Id' and 'UserId', there are no corresponding entries for 37.794204% of entries in 'Id' in 'USERS__STAGING_TABLE_1'. You might want to double-check your join keys.
4	INFO	FOREIGN KEYS NOT FOUND	When joining USERS__STAGING_TABLE_1 and POSTS__STAGING_TABLE_3 over 'Id' and 'OwnerUserId', there are no corresponding entries for 45.600470% of entries in 'Id' in 'USERS__STAGING_TABLE_1'. You might want to double-check your join keys.
5	INFO	FOREIGN KEYS NOT FOUND	When joining POSTS__STAGING_TABLE_3 and VOTES__STAGING_TABLE_4 over 'Id' and 'PostId', there are no corresponding entries for 16.464078% of entries in 'Id' in 'POSTS__STAGING_TABLE_3'. You might want to double-check your join keys.

In [22]:

  Copied!     
 
pipe1.fit(container.train)
pipe1.fit(container.train)

Checking data model...

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00

The pipeline check generated 3 issues labeled INFO and 3 issues labeled WARNING.

To see the issues in full, run .check() on the pipeline.

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  FastProp: Trying 522 features... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:08
  FastProp: Building subfeatures... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  FastProp: Building features... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:01
  XGBoost: Training as predictor... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:09

Trained pipeline.

Time taken: 0:00:19.835015.

Out[22]:

Pipeline(data_model='users',
         feature_learners=['FastProp'],
         feature_selectors=[],
         include_categorical=True,
         loss_function='SquareLoss',
         peripheral=['badges', 'posts', 'votes'],
         predictors=['XGBoostRegressor'],
         preprocessors=[],
         share_selected_features=0.5,
         tags=['fast_prop', 'container-1mZfCR'])

In [23]:

  Copied!     
 
pipe2.check(container.train)
pipe2.check(container.train)

Checking data model...

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Preprocessing... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Checking... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00

The pipeline check generated 3 issues labeled INFO and 3 issues labeled WARNING.

Out[23]:

	type	label	message
0	WARNING	COLUMN SHOULD BE UNUSED	All non-NULL entries in column 't3__posttypeid' in POSTS__STAGING_TABLE_3 are equal to each other. You should consider setting its role to unused_string or using it for comparison only (you can do the latter by setting a unit that contains 'comparison only').
1	WARNING	COLUMN SHOULD BE UNUSED	93.147125% of all entries in column 't3__favoritecount' in POSTS__STAGING_TABLE_3 are NULL values. You should consider setting its role to unused_float.
2	WARNING	COLUMN SHOULD BE UNUSED	99.468396% of all entries in column 'bountyamount' in VOTES__STAGING_TABLE_4 are NULL values. You should consider setting its role to unused_float.
3	INFO	FOREIGN KEYS NOT FOUND	When joining USERS__STAGING_TABLE_1 and BADGES__STAGING_TABLE_2 over 'Id' and 'UserId', there are no corresponding entries for 37.794204% of entries in 'Id' in 'USERS__STAGING_TABLE_1'. You might want to double-check your join keys.
4	INFO	FOREIGN KEYS NOT FOUND	When joining USERS__STAGING_TABLE_1 and POSTS__STAGING_TABLE_3 over 'Id' and 'OwnerUserId', there are no corresponding entries for 45.600470% of entries in 'Id' in 'USERS__STAGING_TABLE_1'. You might want to double-check your join keys.
5	INFO	FOREIGN KEYS NOT FOUND	When joining POSTS__STAGING_TABLE_3 and VOTES__STAGING_TABLE_4 over 'Id' and 'PostId', there are no corresponding entries for 16.464078% of entries in 'Id' in 'POSTS__STAGING_TABLE_3'. You might want to double-check your join keys.

In [24]:

  Copied!     
 
pipe2.fit(container.train)
pipe2.fit(container.train)

Checking data model...

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Preprocessing... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00

The pipeline check generated 3 issues labeled INFO and 3 issues labeled WARNING.

To see the issues in full, run .check() on the pipeline.

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Preprocessing... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Relboost: Training subfeatures... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:19
  Relboost: Training subfeatures... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:22
  Relboost: Building subfeatures... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Relboost: Building subfeatures... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Relboost: Training features... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 01:10
  Relboost: Building subfeatures... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Relboost: Building subfeatures... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Relboost: Building features... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:04
  XGBoost: Training as predictor... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:04

Trained pipeline.

Time taken: 0:02:05.355604.

Out[24]:

Pipeline(data_model='users',
         feature_learners=['Relboost'],
         feature_selectors=[],
         include_categorical=True,
         loss_function='SquareLoss',
         peripheral=['badges', 'posts', 'votes'],
         predictors=['XGBoostRegressor'],
         preprocessors=['Mapping'],
         share_selected_features=0.5,
         tags=['relboost', 'container-1mZfCR'])

2.4 Model evaluation¶

In [25]:

  Copied!     
 
fastprop_score = pipe1.score(container.test)
fastprop_score
fastprop_score = pipe1.score(container.test) fastprop_score

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Preprocessing... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  FastProp: Building subfeatures... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  FastProp: Building features... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00

Out[25]:

	date time	set used	target	mae	rmse	rsquared
0	2024-09-12 12:57:46	train	Reputation	31.8897	43.6771	0.9974
1	2024-09-12 12:59:53	test	Reputation	33.6064	65.3332	0.9777

In [26]:

  Copied!     
 
relboost_score = pipe2.score(container.test)
relboost_score
relboost_score = pipe2.score(container.test) relboost_score

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Preprocessing... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Relboost: Building subfeatures... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Relboost: Building subfeatures... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Relboost: Building features... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:01

Out[26]:

	date time	set used	target	mae	rmse	rsquared
0	2024-09-12 12:59:52	train	Reputation	29.7673	42.9683	0.9975
1	2024-09-12 12:59:55	test	Reputation	32.2748	75.6576	0.9728

2.5 featuretools¶

In [27]:

  Copied!     
 
population_train_pd = container.train.population.to_pandas()
population_test_pd = container.test.population.to_pandas()
population_train_pd = container.train.population.to_pandas() population_test_pd = container.test.population.to_pandas()

In [28]:

  Copied!     
 
badges_pd = badges.drop(badges.roles.unused).to_pandas()
posts_pd = posts.drop(posts.roles.unused).to_pandas()
votes_pd = votes.drop(votes.roles.unused).to_pandas()
badges_pd = badges.drop(badges.roles.unused).to_pandas() posts_pd = posts.drop(posts.roles.unused).to_pandas() votes_pd = votes.drop(votes.roles.unused).to_pandas()

In [29]:

  Copied!     
 
population_pd_logical_types = {
    "Id": ww.logical_types.IntegerNullable,
    "Views": ww.logical_types.Integer,
    "UpVotes": ww.logical_types.Integer,
    "DownVotes": ww.logical_types.Integer,
    "Reputation": ww.logical_types.Integer,
    "CreationDate": ww.logical_types.Datetime,
    "LastAccessDate": ww.logical_types.Datetime
}
population_train_pd.ww.init(logical_types=population_pd_logical_types, index="Id", name="population")
population_test_pd.ww.init(logical_types=population_pd_logical_types, index="Id", name="population")

badges_pd_logical_types = {
    "Name": ww.logical_types.Categorical,
    "Id": ww.logical_types.Integer,
    "UserId": ww.logical_types.IntegerNullable,
    "Date": ww.logical_types.Datetime
}
badges_pd.ww.init(logical_types=badges_pd_logical_types, index="Id", name="badges")

posts_pd_logical_types = {
    "PostTypeId": ww.logical_types.Integer,
    "Id": ww.logical_types.Integer,
    "AcceptedAnswerId": ww.logical_types.IntegerNullable,
    "OwnerUserId": ww.logical_types.IntegerNullable,
    "Score": ww.logical_types.Integer,
    "ViewCount": ww.logical_types.IntegerNullable,
    "AnswerCount": ww.logical_types.IntegerNullable,
    "CommentCount": ww.logical_types.Integer,
    "FavoriteCount": ww.logical_types.IntegerNullable,
    "LastEditorUserId": ww.logical_types.IntegerNullable
}
posts_pd.ww.init(logical_types=posts_pd_logical_types, index="Id", name="posts")

votes_pd_logical_types = {
    "VoteTypeId": ww.logical_types.Categorical,
    "Id": ww.logical_types.Integer,
    "PostId": ww.logical_types.Integer,
    "BountyAmount": ww.logical_types.IntegerNullable,
    "CreationDate": ww.logical_types.Datetime
}
votes_pd.ww.init(logical_types=votes_pd_logical_types, index="Id", name="votes")
population_pd_logical_types = { "Id": ww.logical_types.IntegerNullable, "Views": ww.logical_types.Integer, "UpVotes": ww.logical_types.Integer, "DownVotes": ww.logical_types.Integer, "Reputation": ww.logical_types.Integer, "CreationDate": ww.logical_types.Datetime, "LastAccessDate": ww.logical_types.Datetime } population_train_pd.ww.init(logical_types=population_pd_logical_types, index="Id", name="population") population_test_pd.ww.init(logical_types=population_pd_logical_types, index="Id", name="population") badges_pd_logical_types = { "Name": ww.logical_types.Categorical, "Id": ww.logical_types.Integer, "UserId": ww.logical_types.IntegerNullable, "Date": ww.logical_types.Datetime } badges_pd.ww.init(logical_types=badges_pd_logical_types, index="Id", name="badges") posts_pd_logical_types = { "PostTypeId": ww.logical_types.Integer, "Id": ww.logical_types.Integer, "AcceptedAnswerId": ww.logical_types.IntegerNullable, "OwnerUserId": ww.logical_types.IntegerNullable, "Score": ww.logical_types.Integer, "ViewCount": ww.logical_types.IntegerNullable, "AnswerCount": ww.logical_types.IntegerNullable, "CommentCount": ww.logical_types.Integer, "FavoriteCount": ww.logical_types.IntegerNullable, "LastEditorUserId": ww.logical_types.IntegerNullable } posts_pd.ww.init(logical_types=posts_pd_logical_types, index="Id", name="posts") votes_pd_logical_types = { "VoteTypeId": ww.logical_types.Categorical, "Id": ww.logical_types.Integer, "PostId": ww.logical_types.Integer, "BountyAmount": ww.logical_types.IntegerNullable, "CreationDate": ww.logical_types.Datetime } votes_pd.ww.init(logical_types=votes_pd_logical_types, index="Id", name="votes")

In [30]:

  Copied!     
 
dataframes_train = {
    "population" : (population_train_pd, ),
    "badges": (badges_pd, ),
    "posts": (posts_pd, ),
    "votes": (votes_pd, ),
}
dataframes_train = { "population" : (population_train_pd, ), "badges": (badges_pd, ), "posts": (posts_pd, ), "votes": (votes_pd, ), }

In [31]:

  Copied!     
 
dataframes_test = {
    "population" : (population_test_pd, ),
    "badges": (badges_pd, ),
    "posts": (posts_pd, ),
    "votes": (votes_pd, ),
}
dataframes_test = { "population" : (population_test_pd, ), "badges": (badges_pd, ), "posts": (posts_pd, ), "votes": (votes_pd, ), }

In [32]:

  Copied!     
 
relationships = [
    ("population", "Id", "badges", "UserId"),
    ("population", "Id", "posts", "OwnerUserId"),
    ("posts", "Id", "votes", "PostId"),
]
relationships = [ ("population", "Id", "badges", "UserId"), ("population", "Id", "posts", "OwnerUserId"), ("posts", "Id", "votes", "PostId"), ]

In [33]:

  Copied!     
 
featuretools_train_pd = featuretools.dfs(
    dataframes=dataframes_train,
    relationships=relationships,
    target_dataframe_name="population")[0]
featuretools_train_pd = featuretools.dfs( dataframes=dataframes_train, relationships=relationships, target_dataframe_name="population")[0]

In [34]:

  Copied!     
 
featuretools_test_pd = featuretools.dfs(
    dataframes=dataframes_test,
    relationships=relationships,
    target_dataframe_name="population")[0]
featuretools_test_pd = featuretools.dfs( dataframes=dataframes_test, relationships=relationships, target_dataframe_name="population")[0]

In [35]:

  Copied!     
 
featuretools_train = getml.data.DataFrame.from_pandas(featuretools_train_pd, "featuretools_train")
featuretools_test = getml.data.DataFrame.from_pandas(featuretools_test_pd, "featuretools_test")
featuretools_train = getml.data.DataFrame.from_pandas(featuretools_train_pd, "featuretools_train") featuretools_test = getml.data.DataFrame.from_pandas(featuretools_test_pd, "featuretools_test")

In [36]:

  Copied!     
 
featuretools_train.set_role("Reputation", getml.data.roles.target)
featuretools_train.set_role(featuretools_train.roles.unused_float, getml.data.roles.numerical)
featuretools_train.set_role(featuretools_train.roles.unused_string, getml.data.roles.categorical)
featuretools_train.set_role(["SKEW(posts.STD(votes.BountyAmount))", "STD(posts.SKEW(votes.BountyAmount))"], getml.data.roles.unused_float)

featuretools_train
featuretools_train.set_role("Reputation", getml.data.roles.target) featuretools_train.set_role(featuretools_train.roles.unused_float, getml.data.roles.numerical) featuretools_train.set_role(featuretools_train.roles.unused_string, getml.data.roles.categorical) featuretools_train.set_role(["SKEW(posts.STD(votes.BountyAmount))", "STD(posts.SKEW(votes.BountyAmount))"], getml.data.roles.unused_float) featuretools_train

Out[36]:

name	Reputation	COUNT(badges)	MODE(badges.Name)	NUM_UNIQUE(badges.Name)	COUNT(posts)	COUNT(votes)	MODE(votes.VoteTypeId)	NUM_UNIQUE(votes.VoteTypeId)	DAY(CreationDate)	DAY(LastAccessDate)	MONTH(CreationDate)	MONTH(LastAccessDate)	WEEKDAY(CreationDate)	WEEKDAY(LastAccessDate)	YEAR(CreationDate)	YEAR(LastAccessDate)	MODE(badges.DAY(Date))	MODE(badges.MONTH(Date))	MODE(badges.WEEKDAY(Date))	MODE(badges.YEAR(Date))	NUM_UNIQUE(badges.DAY(Date))	NUM_UNIQUE(badges.MONTH(Date))	NUM_UNIQUE(badges.WEEKDAY(Date))	NUM_UNIQUE(badges.YEAR(Date))	MODE(posts.MODE(votes.VoteTypeId))	NUM_UNIQUE(posts.MODE(votes.VoteTypeId))	Views	UpVotes	DownVotes	MAX(posts.AcceptedAnswerId)	MAX(posts.AnswerCount)	MAX(posts.CommentCount)	MAX(posts.FavoriteCount)	MAX(posts.LastEditorUserId)	MAX(posts.PostTypeId)	MAX(posts.Score)	MAX(posts.ViewCount)	MEAN(posts.AcceptedAnswerId)	MEAN(posts.AnswerCount)	MEAN(posts.CommentCount)	MEAN(posts.FavoriteCount)	MEAN(posts.LastEditorUserId)	MEAN(posts.PostTypeId)	MEAN(posts.Score)	MEAN(posts.ViewCount)	MIN(posts.AcceptedAnswerId)	MIN(posts.AnswerCount)	MIN(posts.CommentCount)	MIN(posts.FavoriteCount)	MIN(posts.LastEditorUserId)	MIN(posts.PostTypeId)	MIN(posts.Score)	MIN(posts.ViewCount)	SKEW(posts.AcceptedAnswerId)	SKEW(posts.AnswerCount)	SKEW(posts.CommentCount)	SKEW(posts.FavoriteCount)	SKEW(posts.LastEditorUserId)	SKEW(posts.PostTypeId)	SKEW(posts.Score)	SKEW(posts.ViewCount)	STD(posts.AcceptedAnswerId)	STD(posts.AnswerCount)	STD(posts.CommentCount)	STD(posts.FavoriteCount)	STD(posts.LastEditorUserId)	STD(posts.PostTypeId)	STD(posts.Score)	STD(posts.ViewCount)	SUM(posts.AcceptedAnswerId)	SUM(posts.AnswerCount)	SUM(posts.CommentCount)	SUM(posts.FavoriteCount)	SUM(posts.LastEditorUserId)	SUM(posts.PostTypeId)	SUM(posts.Score)	SUM(posts.ViewCount)	MAX(votes.BountyAmount)	MEAN(votes.BountyAmount)	MIN(votes.BountyAmount)	SKEW(votes.BountyAmount)	STD(votes.BountyAmount)	SUM(votes.BountyAmount)	MAX(posts.COUNT(votes))	MAX(posts.MEAN(votes.BountyAmount))	MAX(posts.MIN(votes.BountyAmount))	MAX(posts.NUM_UNIQUE(votes.VoteTypeId))	MAX(posts.SKEW(votes.BountyAmount))	MAX(posts.STD(votes.BountyAmount))	MAX(posts.SUM(votes.BountyAmount))	MEAN(posts.COUNT(votes))	MEAN(posts.MAX(votes.BountyAmount))	MEAN(posts.MEAN(votes.BountyAmount))	MEAN(posts.MIN(votes.BountyAmount))	MEAN(posts.NUM_UNIQUE(votes.VoteTypeId))	MEAN(posts.SKEW(votes.BountyAmount))	MEAN(posts.STD(votes.BountyAmount))	MEAN(posts.SUM(votes.BountyAmount))	MIN(posts.COUNT(votes))	MIN(posts.MAX(votes.BountyAmount))	MIN(posts.MEAN(votes.BountyAmount))	MIN(posts.NUM_UNIQUE(votes.VoteTypeId))	MIN(posts.SKEW(votes.BountyAmount))	MIN(posts.STD(votes.BountyAmount))	MIN(posts.SUM(votes.BountyAmount))	SKEW(posts.COUNT(votes))	SKEW(posts.MAX(votes.BountyAmount))	SKEW(posts.MEAN(votes.BountyAmount))	SKEW(posts.MIN(votes.BountyAmount))	SKEW(posts.NUM_UNIQUE(votes.VoteTypeId))	SKEW(posts.SUM(votes.BountyAmount))	STD(posts.COUNT(votes))	STD(posts.MAX(votes.BountyAmount))	STD(posts.MEAN(votes.BountyAmount))	STD(posts.MIN(votes.BountyAmount))	STD(posts.NUM_UNIQUE(votes.VoteTypeId))	STD(posts.SUM(votes.BountyAmount))	SUM(posts.MAX(votes.BountyAmount))	SUM(posts.MEAN(votes.BountyAmount))	SUM(posts.MIN(votes.BountyAmount))	SUM(posts.NUM_UNIQUE(votes.VoteTypeId))	SUM(posts.SKEW(votes.BountyAmount))	SUM(posts.STD(votes.BountyAmount))	MAX(votes.posts.AcceptedAnswerId)	MAX(votes.posts.AnswerCount)	MAX(votes.posts.CommentCount)	MAX(votes.posts.FavoriteCount)	MAX(votes.posts.LastEditorUserId)	MAX(votes.posts.PostTypeId)	MAX(votes.posts.Score)	MAX(votes.posts.ViewCount)	MEAN(votes.posts.AcceptedAnswerId)	MEAN(votes.posts.AnswerCount)	MEAN(votes.posts.CommentCount)	MEAN(votes.posts.FavoriteCount)	MEAN(votes.posts.LastEditorUserId)	MEAN(votes.posts.PostTypeId)	MEAN(votes.posts.Score)	MEAN(votes.posts.ViewCount)	MIN(votes.posts.AcceptedAnswerId)	MIN(votes.posts.AnswerCount)	MIN(votes.posts.CommentCount)	MIN(votes.posts.FavoriteCount)	MIN(votes.posts.LastEditorUserId)	MIN(votes.posts.PostTypeId)	MIN(votes.posts.Score)	MIN(votes.posts.ViewCount)	SKEW(votes.posts.AcceptedAnswerId)	SKEW(votes.posts.AnswerCount)	SKEW(votes.posts.CommentCount)	SKEW(votes.posts.FavoriteCount)	SKEW(votes.posts.LastEditorUserId)	SKEW(votes.posts.PostTypeId)	SKEW(votes.posts.Score)	SKEW(votes.posts.ViewCount)	STD(votes.posts.AcceptedAnswerId)	STD(votes.posts.AnswerCount)	STD(votes.posts.CommentCount)	STD(votes.posts.FavoriteCount)	STD(votes.posts.LastEditorUserId)	STD(votes.posts.PostTypeId)	STD(votes.posts.Score)	STD(votes.posts.ViewCount)	SUM(votes.posts.AcceptedAnswerId)	SUM(votes.posts.AnswerCount)	SUM(votes.posts.CommentCount)	SUM(votes.posts.FavoriteCount)	SUM(votes.posts.LastEditorUserId)	SUM(votes.posts.PostTypeId)	SUM(votes.posts.Score)	SUM(votes.posts.ViewCount)	SKEW(posts.STD(votes.BountyAmount))	STD(posts.SKEW(votes.BountyAmount))
role	target	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	unused_float	unused_float
0	1	0	NULL	NULL	211	10	16	1	19	19	7	7	0	0	2010	2010	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	16	1	0	5007	1920	nan	nan	0	nan	8588	7	0	nan	nan	nan	0	nan	306.1634	4.9005	0	nan	nan	nan	0	nan	-1	3	0	nan	nan	nan	0	nan	4.9461	-0.2648	0	nan	nan	nan	0	nan	1339.1554	0.5386	0	nan	0	0	0	0	61845	1034	0	0	nan	nan	nan	nan	nan	0	1	nan	nan	1	nan	nan	0	0.04739	nan	nan	nan	1	nan	nan	0	0	nan	nan	1	nan	nan	0	4.2908	nan	nan	nan	0	0	0.213	nan	nan	nan	0	0	0	0	0	10	0	0	nan	nan	0	nan	2116	5	0	nan	nan	nan	0	nan	2116	4.4	0	nan	nan	nan	0	nan	2116	3	0	nan	nan	nan	0	nan	nan	-0.7801	0	nan	nan	nan	0	nan	nan	0.6992	0	nan	0	0	0	0	2116	44	0	0	nan	nan
1	101	3	Autobiographer	3	0	0	NULL	NULL	19	12	7	11	0	1	2010	2013	11	7	5	2010	3	3	2	1	NULL	NULL	25	3	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan	nan	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan
2	101	3	Autobiographer	3	0	0	NULL	NULL	19	8	7	8	0	4	2010	2014	2	4	0	2010	3	3	3	2	NULL	NULL	22	19	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan	nan	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan
3	6792	131	Nice Answer	45	117	2016	2	7	19	13	7	8	0	2	2010	2014	19	7	3	2010	30	12	7	5	2	3	1145	662	5	2251	56	7	137	22047	2	156	64481	1793.6667	10.7059	1.5812	29.8125	1004.902	1.8547	12.5043	8898.2941	566	1	0	2	-1	1	1	286	-1.5227	2.6791	1.2464	1.8953	4.868	-2.0393	4.9894	2.5979	616.7514	13.4618	1.7628	45.5869	3514.2488	0.3539	22.4828	17078.287	16143	182	185	477	51250	217	1463	151271	50	50	50	nan	nan	50	303	50	50	4	nan	nan	50	17.2308	50	50	50	1.5299	nan	nan	0.4274	1	50	50	1	nan	nan	0	5.7832	nan	nan	nan	0.9689	10.8167	41.2949	nan	nan	nan	0.6238	4.6225	50	50	50	179	0	0	2251	56	7	137	22047	2	156	64481	1862.9954	25.5357	3.0511	90.5613	5358.5862	1.4648	64.5496	30912.241	566	1	0	2	-1	1	1	286	-1.6226	0.6212	0.3534	-0.561	1.3063	0.1413	0.5721	0.2586	554.8324	20.2126	2.4432	54.3408	8826.2062	0.4989	62.094	23812.9928	402407	27553	6151	97444	7276960	2953	130132	33354308	nan	nan
4	457	19	Announcer	16	12	94	2	3	19	7	7	8	0	3	2010	2014	19	7	2	2010	11	8	5	5	2	2	114	47	0	39226	2	5	4	930	2	40	981	22179	1	1.8333	2.3333	187.3333	1.5	7.0833	507.6667	3832	0	0	1	6	1	0	122	-0.267	0	0.613	0.9352	2.3842	0	2.814	0.2603	14488.6332	0.6325	1.6422	1.5275	366.0419	0.5222	11.0327	399.5571	88716	6	22	7	1124	18	85	3046	nan	nan	nan	nan	nan	0	40	nan	nan	2	nan	nan	0	7.8333	nan	nan	nan	1.3333	nan	nan	0	1	nan	nan	1	nan	nan	0	2.6072	nan	nan	nan	0.8124	0	11.044	nan	nan	nan	0.4924	0	0	0	0	16	0	0	39226	2	5	4	930	2	40	981	23376.875	1.1	1.4681	2.9545	96	1.6809	21.2553	706.7	3832	0	0	1	6	1	0	122	-0.2995	0.8834	0.9674	-0.5825	3.1415	-0.7886	0.1943	-0.8643	13635.7689	0.4026	1.3496	1.3266	245.0523	0.4686	16.5282	334.6668	374030	33	138	65	6240	158	1998	21201	nan	nan
	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
32328	1	0	NULL	NULL	0	0	NULL	NULL	13	13	9	9	5	5	2014	2014	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan	nan	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan
32329	101	1	Supporter	1	0	0	NULL	NULL	13	13	9	9	5	5	2014	2014	13	9	5	2014	1	1	1	1	NULL	NULL	0	1	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan	nan	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan
32330	6	1	Student	1	1	2	2	2	13	13	9	9	5	5	2014	2014	13	9	5	2014	1	1	1	1	2	1	0	0	0	nan	1	0	1	nan	1	1	17	nan	1	0	1	nan	1	1	17	nan	1	0	1	nan	1	1	17	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	1	0	1	0	1	1	17	nan	nan	nan	nan	nan	0	2	nan	nan	2	nan	nan	0	2	nan	nan	nan	2	nan	nan	0	2	nan	nan	2	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	2	0	0	nan	1	0	1	nan	1	1	17	nan	1	0	1	nan	1	1	17	nan	1	0	1	nan	1	1	17	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	nan	0	0	0	0	2	0	2	0	2	2	34	nan	nan
32331	101	0	NULL	NULL	0	0	NULL	NULL	13	13	9	9	5	5	2014	2014	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan	nan	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan
32332	1	0	NULL	NULL	0	0	NULL	NULL	14	14	9	9	6	6	2014	2014	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan	nan	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan

32333 rows x 173 columns
memory usage: 41.52 MB
name: featuretools_train
type: getml.DataFrame

In [37]:

  Copied!     
 
featuretools_test.set_role("Reputation", getml.data.roles.target)
featuretools_test.set_role(featuretools_test.roles.unused_float, getml.data.roles.numerical)
featuretools_test.set_role(featuretools_test.roles.unused_string, getml.data.roles.categorical)
featuretools_test.set_role(["SKEW(posts.STD(votes.BountyAmount))", "STD(posts.SKEW(votes.BountyAmount))"], getml.data.roles.unused_float)

featuretools_test
featuretools_test.set_role("Reputation", getml.data.roles.target) featuretools_test.set_role(featuretools_test.roles.unused_float, getml.data.roles.numerical) featuretools_test.set_role(featuretools_test.roles.unused_string, getml.data.roles.categorical) featuretools_test.set_role(["SKEW(posts.STD(votes.BountyAmount))", "STD(posts.SKEW(votes.BountyAmount))"], getml.data.roles.unused_float) featuretools_test

Out[37]:

name	Reputation	COUNT(badges)	MODE(badges.Name)	NUM_UNIQUE(badges.Name)	COUNT(posts)	COUNT(votes)	MODE(votes.VoteTypeId)	NUM_UNIQUE(votes.VoteTypeId)	DAY(CreationDate)	DAY(LastAccessDate)	MONTH(CreationDate)	MONTH(LastAccessDate)	WEEKDAY(CreationDate)	WEEKDAY(LastAccessDate)	YEAR(CreationDate)	YEAR(LastAccessDate)	MODE(badges.DAY(Date))	MODE(badges.MONTH(Date))	MODE(badges.WEEKDAY(Date))	MODE(badges.YEAR(Date))	NUM_UNIQUE(badges.DAY(Date))	NUM_UNIQUE(badges.MONTH(Date))	NUM_UNIQUE(badges.WEEKDAY(Date))	NUM_UNIQUE(badges.YEAR(Date))	MODE(posts.MODE(votes.VoteTypeId))	NUM_UNIQUE(posts.MODE(votes.VoteTypeId))	Views	UpVotes	DownVotes	MAX(posts.AcceptedAnswerId)	MAX(posts.AnswerCount)	MAX(posts.CommentCount)	MAX(posts.FavoriteCount)	MAX(posts.LastEditorUserId)	MAX(posts.PostTypeId)	MAX(posts.Score)	MAX(posts.ViewCount)	MEAN(posts.AcceptedAnswerId)	MEAN(posts.AnswerCount)	MEAN(posts.CommentCount)	MEAN(posts.FavoriteCount)	MEAN(posts.LastEditorUserId)	MEAN(posts.PostTypeId)	MEAN(posts.Score)	MEAN(posts.ViewCount)	MIN(posts.AcceptedAnswerId)	MIN(posts.AnswerCount)	MIN(posts.CommentCount)	MIN(posts.FavoriteCount)	MIN(posts.LastEditorUserId)	MIN(posts.PostTypeId)	MIN(posts.Score)	MIN(posts.ViewCount)	SKEW(posts.AcceptedAnswerId)	SKEW(posts.AnswerCount)	SKEW(posts.CommentCount)	SKEW(posts.FavoriteCount)	SKEW(posts.LastEditorUserId)	SKEW(posts.PostTypeId)	SKEW(posts.Score)	SKEW(posts.ViewCount)	STD(posts.AcceptedAnswerId)	STD(posts.AnswerCount)	STD(posts.CommentCount)	STD(posts.FavoriteCount)	STD(posts.LastEditorUserId)	STD(posts.PostTypeId)	STD(posts.Score)	STD(posts.ViewCount)	SUM(posts.AcceptedAnswerId)	SUM(posts.AnswerCount)	SUM(posts.CommentCount)	SUM(posts.FavoriteCount)	SUM(posts.LastEditorUserId)	SUM(posts.PostTypeId)	SUM(posts.Score)	SUM(posts.ViewCount)	MAX(votes.BountyAmount)	MEAN(votes.BountyAmount)	MIN(votes.BountyAmount)	SKEW(votes.BountyAmount)	STD(votes.BountyAmount)	SUM(votes.BountyAmount)	MAX(posts.COUNT(votes))	MAX(posts.MEAN(votes.BountyAmount))	MAX(posts.MIN(votes.BountyAmount))	MAX(posts.NUM_UNIQUE(votes.VoteTypeId))	MAX(posts.SKEW(votes.BountyAmount))	MAX(posts.STD(votes.BountyAmount))	MAX(posts.SUM(votes.BountyAmount))	MEAN(posts.COUNT(votes))	MEAN(posts.MAX(votes.BountyAmount))	MEAN(posts.MEAN(votes.BountyAmount))	MEAN(posts.MIN(votes.BountyAmount))	MEAN(posts.NUM_UNIQUE(votes.VoteTypeId))	MEAN(posts.SKEW(votes.BountyAmount))	MEAN(posts.STD(votes.BountyAmount))	MEAN(posts.SUM(votes.BountyAmount))	MIN(posts.COUNT(votes))	MIN(posts.MAX(votes.BountyAmount))	MIN(posts.MEAN(votes.BountyAmount))	MIN(posts.NUM_UNIQUE(votes.VoteTypeId))	MIN(posts.SKEW(votes.BountyAmount))	MIN(posts.STD(votes.BountyAmount))	MIN(posts.SUM(votes.BountyAmount))	SKEW(posts.COUNT(votes))	SKEW(posts.MAX(votes.BountyAmount))	SKEW(posts.MEAN(votes.BountyAmount))	SKEW(posts.MIN(votes.BountyAmount))	SKEW(posts.NUM_UNIQUE(votes.VoteTypeId))	SKEW(posts.SUM(votes.BountyAmount))	STD(posts.COUNT(votes))	STD(posts.MAX(votes.BountyAmount))	STD(posts.MEAN(votes.BountyAmount))	STD(posts.MIN(votes.BountyAmount))	STD(posts.NUM_UNIQUE(votes.VoteTypeId))	STD(posts.SUM(votes.BountyAmount))	SUM(posts.MAX(votes.BountyAmount))	SUM(posts.MEAN(votes.BountyAmount))	SUM(posts.MIN(votes.BountyAmount))	SUM(posts.NUM_UNIQUE(votes.VoteTypeId))	SUM(posts.SKEW(votes.BountyAmount))	SUM(posts.STD(votes.BountyAmount))	MAX(votes.posts.AcceptedAnswerId)	MAX(votes.posts.AnswerCount)	MAX(votes.posts.CommentCount)	MAX(votes.posts.FavoriteCount)	MAX(votes.posts.LastEditorUserId)	MAX(votes.posts.PostTypeId)	MAX(votes.posts.Score)	MAX(votes.posts.ViewCount)	MEAN(votes.posts.AcceptedAnswerId)	MEAN(votes.posts.AnswerCount)	MEAN(votes.posts.CommentCount)	MEAN(votes.posts.FavoriteCount)	MEAN(votes.posts.LastEditorUserId)	MEAN(votes.posts.PostTypeId)	MEAN(votes.posts.Score)	MEAN(votes.posts.ViewCount)	MIN(votes.posts.AcceptedAnswerId)	MIN(votes.posts.AnswerCount)	MIN(votes.posts.CommentCount)	MIN(votes.posts.FavoriteCount)	MIN(votes.posts.LastEditorUserId)	MIN(votes.posts.PostTypeId)	MIN(votes.posts.Score)	MIN(votes.posts.ViewCount)	SKEW(votes.posts.AcceptedAnswerId)	SKEW(votes.posts.AnswerCount)	SKEW(votes.posts.CommentCount)	SKEW(votes.posts.FavoriteCount)	SKEW(votes.posts.LastEditorUserId)	SKEW(votes.posts.PostTypeId)	SKEW(votes.posts.Score)	SKEW(votes.posts.ViewCount)	STD(votes.posts.AcceptedAnswerId)	STD(votes.posts.AnswerCount)	STD(votes.posts.CommentCount)	STD(votes.posts.FavoriteCount)	STD(votes.posts.LastEditorUserId)	STD(votes.posts.PostTypeId)	STD(votes.posts.Score)	STD(votes.posts.ViewCount)	SUM(votes.posts.AcceptedAnswerId)	SUM(votes.posts.AnswerCount)	SUM(votes.posts.CommentCount)	SUM(votes.posts.FavoriteCount)	SUM(votes.posts.LastEditorUserId)	SUM(votes.posts.PostTypeId)	SUM(votes.posts.Score)	SUM(votes.posts.ViewCount)	SKEW(posts.STD(votes.BountyAmount))	STD(posts.SKEW(votes.BountyAmount))
role	target	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	categorical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	numerical	unused_float	unused_float
0	101	2	Autobiographer	2	0	0	NULL	NULL	19	2	7	1	0	3	2010	2014	8	6	5	2011	2	2	1	2	NULL	NULL	11	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan	nan	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan
1	429	10	Yearling	9	2	59	2	4	19	10	7	9	0	2	2010	2014	19	7	2	2010	8	6	5	4	2	1	56	20	0	nan	7	4	7	17230	2	26	1582	nan	7	3	7	8659	1.5	23.5	1582	nan	7	2	7	88	1	21	1582	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	1.4142	nan	12121.2244	0.7071	3.5355	nan	0	7	6	7	17318	3	47	1582	nan	nan	nan	nan	nan	0	30	nan	nan	3	nan	nan	0	29.5	nan	nan	nan	3	nan	nan	0	29	nan	nan	3	nan	nan	0	nan	nan	nan	nan	nan	nan	0.7071	nan	nan	nan	0	0	0	0	0	6	0	0	nan	7	4	7	17230	2	26	1582	nan	7	2.9831	7	8513.7288	1.4915	23.4576	1582	nan	7	2	7	88	1	21	1582	nan	0	0.03479	0	0.03479	0.03479	0.03479	0	nan	0	1.0084	0	8643.3304	0.5042	2.5211	0	0	210	176	210	502310	88	1384	47460	nan	nan
2	101	5	Autobiographer	5	1	2	2	1	19	9	7	9	0	1	2010	2014	19	7	0	2010	4	3	4	2	2	1	10	5	0	nan	nan	0	nan	615	2	2	nan	nan	nan	0	nan	615	2	2	nan	nan	nan	0	nan	615	2	2	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	615	2	2	0	nan	nan	nan	nan	nan	0	2	nan	nan	1	nan	nan	0	2	nan	nan	nan	1	nan	nan	0	2	nan	nan	1	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	1	0	0	nan	nan	0	nan	615	2	2	nan	nan	nan	0	nan	615	2	2	nan	nan	nan	0	nan	615	2	2	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	nan	0	0	0	nan	0	0	0	0	1230	4	4	0	nan	nan
3	2185	44	Announcer	20	39	499	2	5	19	27	7	8	0	2	2010	2014	10	7	1	2012	23	12	7	5	2	2	312	40	3	15396	11	10	21	28666	2	74	1926	7502.75	2.4	2.1282	5.7692	3279.1053	1.4872	10.5641	655.75	844	0	0	1	30	1	0	73	0.3969	2.2973	1.2348	1.4673	2.7866	0.05337	3.4381	1.0282	4550.3906	2.5215	2.4299	7.0018	7949.6098	0.5064	15.5306	588.8215	60022	48	83	75	62303	58	412	13115	nan	nan	nan	nan	nan	0	74	nan	nan	4	nan	nan	0	12.7949	nan	nan	nan	1.5789	nan	nan	0	0	nan	nan	1	nan	nan	0	2.6117	nan	nan	nan	1.3039	0	17.2095	nan	nan	nan	0.7215	0	0	0	0	60	0	0	15396	11	10	21	28666	2	74	1926	6143.6699	4.4669	2.6934	11.1415	8242.8584	1.515	30.1603	1191.1488	844	0	0	1	30	1	1	73	0.5685	0.9218	0.7248	-0.0435	0.8061	-0.06033	0.7466	-0.3801	3575.4252	3.6326	2.4585	8.0147	11448.7075	0.5003	27.7176	636.2425	632798	1081	1344	2362	2736629	756	15050	288258	nan	nan
4	1651	22	Nice Answer	14	14	172	2	4	19	7	7	8	0	3	2010	2014	19	7	0	2010	12	9	7	5	2	1	255	12	0	nan	4	5	6	805	2	37	413	nan	2.5	1.7857	4.5	102.8333	1.8571	11.1429	376	nan	1	0	3	39	1	1	339	nan	nan	0.8296	nan	3.4641	-2.2948	1.703	nan	nan	2.1213	1.8051	2.1213	221.1252	0.3631	9.5905	52.3259	0	5	25	9	1234	26	156	752	nan	nan	nan	nan	nan	0	37	nan	nan	3	nan	nan	0	12.2857	nan	nan	nan	1.6429	nan	nan	0	1	nan	nan	1	nan	nan	0	1.3957	nan	nan	nan	0.4326	0	9.7067	nan	nan	nan	0.6333	0	0	0	0	23	0	0	nan	4	5	6	805	2	37	413	nan	2.4483	1.9419	4.5517	177.3056	1.8314	18.0756	377.2759	nan	1	0	3	39	1	1	339	nan	0.07283	0.7614	-0.07283	1.6785	-1.7859	0.6603	-0.07283	nan	1.5256	1.6533	1.5256	295.6699	0.3755	11.4253	37.6325	0	71	334	132	25532	315	3109	10941	nan	nan
	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
7987	1	0	NULL	NULL	0	0	NULL	NULL	13	13	9	9	5	5	2014	2014	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan	nan	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan
7988	11	1	Student	1	1	2	2	1	13	14	9	9	5	6	2014	2014	13	9	5	2014	1	1	1	1	2	1	0	0	0	nan	0	4	nan	nan	1	2	40	nan	0	4	nan	nan	1	2	40	nan	0	4	nan	nan	1	2	40	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	4	0	0	1	2	40	nan	nan	nan	nan	nan	0	2	nan	nan	1	nan	nan	0	2	nan	nan	nan	1	nan	nan	0	2	nan	nan	1	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	1	0	0	nan	0	4	nan	nan	1	2	40	nan	0	4	nan	nan	1	2	40	nan	0	4	nan	nan	1	2	40	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	nan	nan	0	0	0	0	0	8	0	0	2	4	80	nan	nan
7989	1	0	NULL	NULL	0	0	NULL	NULL	13	13	9	9	5	5	2014	2014	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan	nan	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	0	0	0	0	0	nan	nan
7990	6	1	Student	1	1	1	2	1	13	13	9	9	5	5	2014	2014	13	9	5	2014	1	1	1	1	2	1	1	0	0	nan	0	2	nan	nan	1	1	13	nan	0	2	nan	nan	1	1	13	nan	0	2	nan	nan	1	1	13	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	2	0	0	1	1	13	nan	nan	nan	nan	nan	0	1	nan	nan	1	nan	nan	0	1	nan	nan	nan	1	nan	nan	0	1	nan	nan	1	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	1	0	0	nan	0	2	nan	nan	1	1	13	nan	0	2	nan	nan	1	1	13	nan	0	2	nan	nan	1	1	13	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	2	0	0	1	1	13	nan	nan
7991	106	1	Student	1	1	1	2	1	14	14	9	9	6	6	2014	2014	14	9	6	2014	1	1	1	1	2	1	1	0	0	nan	0	2	nan	7290	1	1	5	nan	0	2	nan	7290	1	1	5	nan	0	2	nan	7290	1	1	5	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	2	0	7290	1	1	5	nan	nan	nan	nan	nan	0	1	nan	nan	1	nan	nan	0	1	nan	nan	nan	1	nan	nan	0	1	nan	nan	1	nan	nan	0	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	0	1	0	0	nan	0	2	nan	7290	1	1	5	nan	0	2	nan	7290	1	1	5	nan	0	2	nan	7290	1	1	5	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	nan	0	0	2	0	7290	1	1	5	nan	nan

7992 rows x 173 columns
memory usage: 10.26 MB
name: featuretools_test
type: getml.DataFrame

We train an untuned XGBoostRegressor on top of featuretools' features, just like we have done for getML's features.

Since some of featuretools features are categorical, we allow the pipeline to include these features as well. Other features contain NaN values, which is why we also apply getML's Imputation preprocessor.

In [38]:

  Copied!     
 
imputation = getml.preprocessors.Imputation()

predictor = getml.predictors.XGBoostRegressor(n_jobs=1)

pipe3 = getml.Pipeline(
    tags=['featuretools'],
    preprocessors=[imputation],
    predictors=[predictor],
    include_categorical=True,
)

pipe3
imputation = getml.preprocessors.Imputation() predictor = getml.predictors.XGBoostRegressor(n_jobs=1) pipe3 = getml.Pipeline( tags=['featuretools'], preprocessors=[imputation], predictors=[predictor], include_categorical=True, ) pipe3

Out[38]:

Pipeline(data_model='population',
         feature_learners=[],
         feature_selectors=[],
         include_categorical=True,
         loss_function='SquareLoss',
         peripheral=[],
         predictors=['XGBoostRegressor'],
         preprocessors=['Imputation'],
         share_selected_features=0.5,
         tags=['featuretools'])

In [39]:

  Copied!     
 
pipe3.fit(featuretools_train)
pipe3.fit(featuretools_train)

Checking data model...

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Preprocessing... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Checking... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00

OK.

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Preprocessing... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  XGBoost: Training as predictor... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:08

Trained pipeline.

Time taken: 0:00:08.607237.

Out[39]:

Pipeline(data_model='population',
         feature_learners=[],
         feature_selectors=[],
         include_categorical=True,
         loss_function='SquareLoss',
         peripheral=[],
         predictors=['XGBoostRegressor'],
         preprocessors=['Imputation'],
         share_selected_features=0.5,
         tags=['featuretools'])

In [40]:

  Copied!     
 
featuretools_score = pipe3.score(featuretools_test)
featuretools_score
featuretools_score = pipe3.score(featuretools_test) featuretools_score

  Staging... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00
  Preprocessing... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% • 00:00

Out[40]:

	date time	set used	target	mae	rmse	rsquared
0	2024-09-12 13:00:56	featuretools_train	Reputation	32.3818	48.2617	0.9969
1	2024-09-12 13:00:57	featuretools_test	Reputation	34.219	78.0558	0.9684

2.6 Features¶

The most important features look as follows:

In [41]:

  Copied!     
 
pipe1.features.to_sql()[pipe1.features.sort(by="importances")[0].name]
pipe1.features.to_sql()[pipe1.features.sort(by="importances")[0].name]

Out[41]:

DROP TABLE IF EXISTS "FEATURE_1_87";

CREATE TABLE "FEATURE_1_87" AS
SELECT SUM( t2."score" ) AS "feature_1_87",
       t1.rowid AS rownum
FROM "USERS__STAGING_TABLE_1" t1
INNER JOIN "POSTS__STAGING_TABLE_3" t2
ON t1."id" = t2."owneruserid"
WHERE t2."posttypeid" = '2'
GROUP BY t1.rowid;

In [42]:

  Copied!     
 
pipe2.features.to_sql()[pipe2.features.sort(by="importances")[0].name]
pipe2.features.to_sql()[pipe2.features.sort(by="importances")[0].name]

Out[42]:

DROP TABLE IF EXISTS "FEATURE_1_1";

CREATE TABLE "FEATURE_1_1" AS
SELECT SUM( 
    CASE
        WHEN ( f_1_2."feature_1_2_13" > 44.522517 ) AND ( f_1_2."feature_1_2_12" > 1114.704894 ) THEN 883.3731803463131
        WHEN ( f_1_2."feature_1_2_13" > 44.522517 ) AND ( f_1_2."feature_1_2_12" <= 1114.704894 ) AND ( t2."t3__answercount" > 1.000000 ) THEN 180.6530491037182
        WHEN ( f_1_2."feature_1_2_13" > 44.522517 ) AND ( f_1_2."feature_1_2_12" <= 1114.704894 ) AND ( t2."t3__answercount" <= 1.000000 OR t2."t3__answercount" IS NULL ) THEN 117.0240683994715
        WHEN ( f_1_2."feature_1_2_13" <= 44.522517 ) AND ( f_1_2."feature_1_2_16" > 6.908843 ) AND ( t1."upvotes" > 2258.000000 ) THEN 12.25445123010827
        WHEN ( f_1_2."feature_1_2_13" <= 44.522517 ) AND ( f_1_2."feature_1_2_16" > 6.908843 ) AND ( t1."upvotes" <= 2258.000000 OR t1."upvotes" IS NULL ) THEN 14.16101412804239
        WHEN ( f_1_2."feature_1_2_13" <= 44.522517 ) AND ( f_1_2."feature_1_2_16" <= 6.908843 ) AND ( t1."upvotes__mapping_target_1_avg" > 90.840760 ) THEN 6.041694975904914
        WHEN ( f_1_2."feature_1_2_13" <= 44.522517 ) AND ( f_1_2."feature_1_2_16" <= 6.908843 ) AND ( t1."upvotes__mapping_target_1_avg" <= 90.840760 OR t1."upvotes__mapping_target_1_avg" IS NULL ) THEN -1.015010571805766
        ELSE NULL
    END
) AS "feature_1_1",
       t1.rowid AS rownum
FROM "USERS__STAGING_TABLE_1" t1
INNER JOIN "POSTS__STAGING_TABLE_3" t2
ON t1."id" = t2."owneruserid"
LEFT JOIN "FEATURES_1_2" f_1_2
ON t2.rowid = f_1_2."rownum"
GROUP BY t1.rowid;

2.7 Productionization¶

It is possible to productionize the pipeline by transpiling the features into production-ready SQL code. Please also refer to getML's sqlite3 and spark modules.

In [ ]:

  Copied!     
 
# Creates a folder named stats_pipeline containing
# the SQL code.
pipe2.features.to_sql().save("stats_pipeline")
# Creates a folder named stats_pipeline containing # the SQL code. pipe2.features.to_sql().save("stats_pipeline")

In [ ]:

  Copied!     
 
pipe2.features.to_sql(dialect=getml.pipeline.dialect.spark_sql).save("stats_spark")
pipe2.features.to_sql(dialect=getml.pipeline.dialect.spark_sql).save("stats_spark")

2.8 Discussion¶

For a more convenient overview, we summarize our results into a table.

In [45]:

  Copied!     
 
scores = [fastprop_score, relboost_score, featuretools_score]
pd.DataFrame(data={
    'Name': ['getML: FastProp', 'getML: Relboost', 'featuretools'],
    'R-squared': [f'{score.rsquared:.2%}' for score in scores],
    'RMSE': [f'{score.rmse:,.2f}' for score in scores],
    'MAE': [f'{score.mae:,.2f}' for score in scores]
})
scores = [fastprop_score, relboost_score, featuretools_score] pd.DataFrame(data={ 'Name': ['getML: FastProp', 'getML: Relboost', 'featuretools'], 'R-squared': [f'{score.rsquared:.2%}' for score in scores], 'RMSE': [f'{score.rmse:,.2f}' for score in scores], 'MAE': [f'{score.mae:,.2f}' for score in scores] })

Out[45]:

	Name	R-squared	RMSE	MAE
0	getML: FastProp	97.77%	65.33	33.61
1	getML: Relboost	97.28%	75.66	32.27
2	featuretools	96.84%	78.06	34.22

In [ ]:

  Copied!     
 
getml.engine.shutdown()
getml.engine.shutdown()

3. Conclusion¶

In this notebook, we have benchmarked several relational learning algorithms based on their ability to reverse-engineer StatsExchange's system of calculating users' reputation. We have found that all relational learning algorithm can reverse-engineer the system with high levels of accuracy, but we have also found that getML nonetheless outperforms featuretools.

References¶

Motl, Jan, and Oliver Schulte. "The CTU prague relational learning repository." arXiv preprint arXiv:1511.03086 (2015).