Frequently Asked Questions

Here we are attempting to answer some commonly asked questions that appear on Github, and Stack Overflow.

[1]:
import featuretools as ft
import pandas as pd
import numpy as np

EntitySet

How do I get a list of variable (column) names, and types in an EntitySet?

After you create your EntitySet, you may wish to view the column names. An EntitySet contains multiple Dataframes, one for each entity.

[ ]:
es = ft.demo.load_mock_customer(return_entityset=True)
es

If you want view the variables (columns), and types for the “transactions” entity, you can do the following:

[ ]:
es['transactions'].variables

If you want to view the underlying Dataframe, you can do the following:

[ ]:
es['transactions'].df.head()

What is the difference between copy_variables and additional_variables?

The function normalize_entity creates a new entity and a relationship from unique values of an existing entity. It takes 2 similar arguments:

  • additional_variables removes variables from the base entity and moves them to the new entity.

  • copy_variables keeps the given variables in the base entity, but also copies them to the new entity.

[ ]:
data = ft.demo.load_mock_customer()
transactions_df = data["transactions"].merge(data["sessions"]).merge(data["customers"])
products_df = data["products"]

es = ft.EntitySet(id="customer_data")
es = es.entity_from_dataframe(entity_id="transactions",
                              dataframe=transactions_df,
                              index="transaction_id",
                              time_index="transaction_time")

es = es.entity_from_dataframe(entity_id="products",
                              dataframe=products_df,
                              index="product_id")

new_relationship = ft.Relationship(es["products"]["product_id"], es["transactions"]["product_id"])
es = es.add_relationship(new_relationship)

Before we normalize to create a new entity, let’s look at base entity

[ ]:
es['transactions'].df.head()

Notice the columns session_id, session_start, join_date, device, customer_id, and zip_code.

[ ]:
es = es.normalize_entity(base_entity_id="transactions",
                         new_entity_id="sessions",
                         index="session_id",
                         make_time_index="session_start",
                         additional_variables=["session_start", "join_date"],
                         copy_variables=["device", "customer_id", "zip_code"])

Above, we normalized the columns to create a new entity. - For additional_variables, the following columns ['session_start', 'join_date] will be removed from the products entity, and moved to the new device entity.

  • For copy_variables, the following columns ['device', 'customer_id', 'zip_code'] will be copied from the products entity to the new device entity.

Let’s see this in the actual EntitySet.

[ ]:
es['transactions'].df.head()

Notice above how ['device', 'customer_id', 'zip_code'] are still in the transactions entity, while ['session_start', 'join_date'] are not. But, they have all been moved to the sessions entity, as seen below.

[ ]:
es['sessions'].df.head()

Why did variable type change to Id, Index, or datetime_time_index?

During the creation of your EntitySet, you might be wondering why your variable type changed.

[ ]:
data = ft.demo.load_mock_customer()
transactions_df = data["transactions"].merge(data["sessions"]).merge(data["customers"])
products_df = data["products"]

es = ft.EntitySet(id="customer_data")
es = es.entity_from_dataframe(entity_id="transactions",
                              dataframe=transactions_df,
                              index="transaction_id",
                              time_index="transaction_time")
es.plot()

Notice how the variable type of session_id is Numeric, and the variable type of session_start is Datetime.

Now, let’s normalize the transactions entity to create a new entity.

[ ]:
es = es.normalize_entity(base_entity_id="transactions",
                         new_entity_id="sessions",
                         index="session_id",
                         make_time_index="session_start",
                         additional_variables=["session_start"])
es.plot()

The type for session_id is now Id in the transactions entity, and Index in the new entity, sessions. This is the case because when we normalize the entity, we create a new relationship between the transactions and sessions. There is a one to many relationship between the parent entity, sessions, and child entity, transactions.

Therefore, session_id has type Id in transactions because it represents an Index in another entity. There would be a similar effect if we added another entity using entity_from_dataframe and add_relationship.

In addition, when we created the new entity, we specified a time_index which was the variable (column) session_start. This changed the type of session_start to datetime_time_index in the new sessions entity because it now represents a time_index.

How do I combine two or more interesting values?

You might want to create features that are conditioned on multiple values before they are calculated. This would require the use of interesting_values. However, since we are trying to create the feature with multiple conditions, we will need to modify the Dataframe before we create the EntitySet.

Let’s look at how you might accomplish this.

First, let’s create our Dataframes.

[ ]:
data = ft.demo.load_mock_customer()
transactions_df = data["transactions"].merge(data["sessions"]).merge(data["customers"])
products_df = data["products"]
[ ]:
transactions_df.head()
[ ]:
products_df.head()

Now, let’s modify our transactions Dataframe to create the additional column that represents multiple conditions for our feature.

[ ]:
transactions_df['product_id_device'] = transactions_df['product_id'].astype(str) + ' and ' + transactions_df['device']

Here, we created a new column called product_id_device, which just combines the product_id column, and the device column.

Now let’s create our EntitySet.

[ ]:
es = ft.EntitySet(id="customer_data")
es = es.entity_from_dataframe(entity_id="transactions",
                              dataframe=transactions_df,
                              index="transaction_id",
                              time_index="transaction_time",
                              variable_types={"product_id": ft.variable_types.Categorical,
                                              "product_id_device": ft.variable_types.Categorical,
                                              "zip_code": ft.variable_types.ZIPCode})

es = es.entity_from_dataframe(entity_id="products",
                              dataframe=products_df,
                              index="product_id")
es = es.normalize_entity(base_entity_id="transactions",
                         new_entity_id="sessions",
                         index="session_id",
                         additional_variables=["device", "product_id_device", "customer_id"])
es = es.normalize_entity(base_entity_id="sessions",
                         new_entity_id="customers",
                         index="customer_id")
es

Now, we are ready to add our interesting values.

First, let’s view our options for what the interesting values could be.

[ ]:
interesting_values = transactions_df['product_id_device'].unique().tolist()
interesting_values

If you wanted to, you could pick a subset of these, and the where features created would only use those conditions. In our example, we will use all the possible interesting values.

Here, we set all of these values as our interesting values for this specific entity and variable. If we wanted to, we could make interesting values in the same way for more than one variable, but we will just stick with this one for this example.

[ ]:
es['sessions']['product_id_device'].interesting_values = interesting_values

Now we can run DFS.

[ ]:
feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="customers",
                                      agg_primitives=["count"],
                                      where_primitives=["count"],
                                      trans_primitives=[])
feature_matrix.head()

To better understand the where clause features, let’s examine one of those features. The feature COUNT(sessions WHERE product_id_device = 5 and tablet), tells us how many sessions the customer purchased product_id 5 while on a tablet. Notice how the feature depends on multiple conditions (product_id = 5 & device = tablet).

[ ]:
feature_matrix[["COUNT(sessions WHERE product_id_device = 5 and tablet)"]]

DFS

Why is DFS not creating aggregation features?

You may have created your EntitySet, and then applied DFS to create features. However, you may be puzzled as to why no aggregation features were created.

  • This is most likely because you have a single table in your entity, and DFS is not capable of creating aggregation features with fewer than 2 entities. Featuretools looks for a relationship, and aggregates based on that relationship.

Let’s look at a simple example.

[ ]:
data = ft.demo.load_mock_customer()
transactions_df = data["transactions"].merge(data["sessions"]).merge(data["customers"])

es = ft.EntitySet(id="customer_data")
es = es.entity_from_dataframe(entity_id="transactions",
                              dataframe=transactions_df,
                              index="transaction_id")
es

Notice how we only have 1 entity in our EntitySet. If we try to create aggregation features on this EntitySet, it will not be possible because DFS needs 2 entities to generate aggregation features.

[ ]:
feature_matrix, feature_defs = ft.dfs(entityset=es, target_entity="transactions")
feature_defs

None of the above features are aggregation features. To fix this issue, you can add another entity to your EntitySet.

Solution #1 - You can add new entity if you have additional data.

[ ]:
products_df = data["products"]
es = es.entity_from_dataframe(entity_id="products",
                              dataframe=products_df,
                              index="product_id")
es

Notice how we now have an additional entity in our EntitySet, called products.

Solution #2 - You can normalize an existing entity.

[ ]:
es = es.normalize_entity(base_entity_id="transactions",
                         new_entity_id="sessions",
                         index="session_id",
                         make_time_index="session_start",
                         additional_variables=["device", "customer_id", "zip_code", "session_start", "join_date"])
es

Notice how we now have an additional entity in our EntitySet, called sessions. Here, the normalization created a relationship between transactions and sessions. However, we could have specified a relationship between transactions and products if we had only used Solution #1.

Now, we can generate aggregation features.

[ ]:
feature_matrix, feature_defs = ft.dfs(entityset=es, target_entity="transactions")
feature_defs[:-10]

A few of the aggregation features are:

  • <Feature: sessions.SUM(transactions.amount)>

  • <Feature: sessions.STD(transactions.amount)>

  • <Feature: sessions.MAX(transactions.amount)>

  • <Feature: sessions.SKEW(transactions.amount)>

  • <Feature: sessions.MIN(transactions.amount)>

  • <Feature: sessions.MEAN(transactions.amount)>

  • <Feature: sessions.COUNT(transactions)>

How do I speed up the runtime of DFS?

One issue you may encounter while running ft.dfs is slow performance. While Featuretools has generally optimal default settings for calculating features, you may want to speed up performance when you are calculating on a large number of features.

One quick way to speed up performance is by adjusting the n_jobs settings of ft.dfs or ft.calculate_feature_matrix.

# setting n_jobs to -1 will use all cores

feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="customers",
                                      n_jobs=-1)


feature_matrix, feature_defs = ft.calculate_feature_matrix(entityset=es,
                                                           features=feature_defs,
                                                           n_jobs=-1)

For more ways to speed up performance, please visit:

How do I include only certain features when running DFS?

When using DFS to generate features, you may wish to include only certain features. There are multiple ways that you do this:

  • Use the ignore_variables to specify variables in an entity that should not be used to create features. It is a dictionary mapping an entity id to a list of variable names to ignore.

  • Use drop_contains to drop features that contain any of the strings listed in this parameter.

  • Use drop_exact to drop features that exactly match any of the strings listed in this parameter.

Here is an example of using all three parameters:

[ ]:
es = ft.demo.load_mock_customer(return_entityset=True)

feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="customers",
                                      ignore_variables={
                                           "transactions": ["amount"],
                                           "customers": ["age", "gender", "date_of_birth"]
                                       }, # ignore these variables
                                      drop_contains=["customers.SUM("], # drop features that contain these strings
                                      drop_exact=["STD(transactions.quanity)"]) # drop features that exactly match

If I didn’t specify the cutoff_time, what date will be used for the feature calculations?

The cutoff time will be set to the current time using cutoff_time = datetime.now().

How do I select a certain amount of past data when calculating features?

You may encounter a situation when you wish to make prediction using only a certain amount of historical data. You can accomplish this using the training_window parameter in ft.dfs. When you use the training_window, Featuretools will use the historical data between the cutoff_time and cutoff_time - training_window.

In order to make the calculation, Featuretools will check the time in the time_index column of the target_entity.

[ ]:
es = ft.demo.load_mock_customer(return_entityset=True)
es['customers'].time_index

Our target_entity has a time_index, which is needed for the training_window calculation. Here, we are creating a cutoff time dataframe so that we can have a unique training window for each customer.

[ ]:
cutoff_times = pd.DataFrame()
cutoff_times['customer_id'] = [1, 2, 3, 1]
cutoff_times['time'] = pd.to_datetime(['2014-1-1 04:00', '2014-1-1 05:00', '2014-1-1 06:00', '2014-1-1 08:00'])
cutoff_times['label'] = [True, True, False, True]

feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="customers",
                                      cutoff_time=cutoff_times,
                                      cutoff_time_in_index=True,
                                      training_window="1 hour")
feature_matrix.head()

Above, we ran DFS with training_window argument of 1 hour to create features that only used customer data collected in the last hour (from the cutoff time we provided).

How do I apply DFS to a single table?

You can run DFS on a single table. Featuretools will be able to generate features for your data, but only transform features.

For example:

[2]:
transactions_df = ft.demo.load_mock_customer(return_single_table=True)

es = ft.EntitySet(id="customer_data")
es = es.entity_from_dataframe(entity_id="transactions",
                              dataframe=transactions_df,
                              index="transaction_id",
                              time_index="transaction_time")

feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="transactions",
                                      trans_primitives=['time_since', 'day', 'is_weekend',
                                                        'cum_min', 'minute',
                                                        'num_words', 'weekday', 'cum_count',
                                                        'percentile', 'year', 'week',
                                                        'cum_mean'])

Before we examine the output, let’s look at our original single table.

[4]:
transactions_df.head()
[4]:
transaction_id session_id transaction_time product_id amount customer_id device session_start zip_code join_date date_of_birth brand
0 298 1 2014-01-01 00:00:00 5 127.64 2 desktop 2014-01-01 00:00:00 13244 2012-04-15 23:31:04 1986-08-18 A
1 10 1 2014-01-01 00:09:45 5 57.39 2 desktop 2014-01-01 00:00:00 13244 2012-04-15 23:31:04 1986-08-18 A
2 495 1 2014-01-01 00:14:05 5 69.45 2 desktop 2014-01-01 00:00:00 13244 2012-04-15 23:31:04 1986-08-18 A
3 460 10 2014-01-01 02:33:50 5 123.19 2 tablet 2014-01-01 02:31:40 13244 2012-04-15 23:31:04 1986-08-18 A
4 302 10 2014-01-01 02:37:05 5 64.47 2 tablet 2014-01-01 02:31:40 13244 2012-04-15 23:31:04 1986-08-18 A

Now we can look at the transformations that Featuretools was able to apply to this single entity (table) to create feature matrix.

[5]:
feature_matrix.head()
[5]:
session_id product_id amount customer_id device zip_code brand DAY(transaction_time) DAY(session_start) DAY(join_date) ... CUM_MEAN(CUM_MIN(session_id)) CUM_MEAN(CUM_MIN(amount)) CUM_MEAN(CUM_MIN(customer_id)) CUM_MEAN(MINUTE(transaction_time)) CUM_MEAN(MINUTE(session_start)) CUM_MEAN(MINUTE(join_date)) CUM_MEAN(MINUTE(date_of_birth)) CUM_MEAN(PERCENTILE(session_id)) CUM_MEAN(PERCENTILE(amount)) CUM_MEAN(PERCENTILE(customer_id))
transaction_id
1 31 2 21.77 2 mobile 13244 B 1 1 15 ... 1.0 7.511166 1.095571 28.720280 28.622378 31.000000 0.0 0.430075 0.507935 0.501170
2 1 2 109.48 2 desktop 13244 B 1 1 15 ... 1.0 118.560000 2.000000 0.500000 0.000000 31.000000 0.0 0.017000 0.793000 0.346000
3 35 3 62.49 3 mobile 13244 B 1 1 13 ... 1.0 7.305505 1.084536 28.886598 27.985567 31.364948 0.0 0.486031 0.504231 0.500041
4 30 3 7.55 5 desktop 60091 B 1 1 17 ... 1.0 7.558038 1.098086 28.480861 28.593301 31.086124 0.0 0.419108 0.505258 0.492852
5 4 2 126.80 1 mobile 60091 B 1 1 17 ... 1.0 16.696667 1.621212 25.681818 25.606061 31.606061 0.0 0.067000 0.496621 0.438500

5 rows × 57 columns

Can I automatically normalize a single table?

Yes, another open source library AutoNormalize, also produced by Feature Labs, automates table normalization and integrates with Featuretools. To install run:

python -m pip install featuretools[autonormalize]

A normalized EntitySet will help Featuretools to generate more features. For example:

[3]:
from featuretools.autonormalize import autonormalize as an
es = an.normalize_entity(es)
es.plot()
100%|██████████| 10/10 [00:03<00:00,  3.18it/s]
[3]:
_images/frequently_asked_questions_68_1.svg

As you can see, AutoNormalize creates a relational EntitySet. Below, we run dfs on the EntitySet, and you can see all the features created; take note of the aggregated features.

[4]:
feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="transaction_id",
                                      trans_primitives=[])
feature_matrix.head()
[4]:
session_id product_id amount session_id.customer_id session_id.device product_id.brand session_id.SUM(transaction_id.amount) session_id.STD(transaction_id.amount) session_id.MAX(transaction_id.amount) session_id.SKEW(transaction_id.amount) ... session_id.customer_id.zip_code product_id.SUM(transaction_id.amount) product_id.STD(transaction_id.amount) product_id.MAX(transaction_id.amount) product_id.SKEW(transaction_id.amount) product_id.MIN(transaction_id.amount) product_id.MEAN(transaction_id.amount) product_id.COUNT(transaction_id) product_id.NUM_UNIQUE(transaction_id.session_id) product_id.MODE(transaction_id.session_id)
transaction_id
1 31 2 21.77 2 mobile B 1240.19 39.414339 143.45 0.755711 ... 13244 7021.43 46.336308 149.95 0.151934 5.73 76.319891 92 34 28
2 1 2 109.48 2 desktop B 1229.01 41.600976 141.66 0.295458 ... 13244 7021.43 46.336308 149.95 0.151934 5.73 76.319891 92 34 28
3 35 3 62.49 3 mobile B 960.67 41.880342 148.31 0.854976 ... 13244 7008.12 38.871405 148.31 0.223938 5.89 73.001250 96 35 1
4 30 3 7.55 5 desktop B 1221.13 39.658094 144.53 -0.200571 ... 60091 7008.12 38.871405 148.31 0.223938 5.89 73.001250 96 35 1
5 4 2 126.80 1 mobile B 1613.93 40.187205 129.00 0.234349 ... 60091 7021.43 46.336308 149.95 0.151934 5.73 76.319891 92 34 28

5 rows × 25 columns

How do I prevent label leakage with DFS?

One concern you might have with using DFS is about label leakage. You want to make sure that labels in your data aren’t used incorrectly to create features and the feature matrix.

Featuretools is particularly focused on helping users avoid label leakage.

There are two ways to prevent label leakage depending on if your data has timestamps or not.

1. Data without timestamps

In the case where you do not have timestamps, you can create one EntitySet using only the training data and then run ft.dfs. This will create a feature matrix using only the training data, but also return a list of feature definitions. Next, you can create an EntitySet using the test data and recalculate the same features by calling ft.calculate_feature_matrix with the list of feature definitions from before.

Here is what that flow would look like:

First, let’s create our training data.

[ ]:
train_data = pd.DataFrame({"customer_id": [1, 2, 3, 4, 5],
                           "age": [40, 50, 10, 20, 30],
                           "gender": ["m", "f", "m", "f", "f"],
                           "signup_date": pd.date_range('2014-01-01 01:41:50', periods=5, freq='25min'),
                           "labels": [True, False, True, False, True]})
train_data.head()

Now, we can create an entityset for our training data.

[ ]:
es_train_data = ft.EntitySet(id="customer_train_data")
es_train_data = es_train_data.entity_from_dataframe(entity_id="customers",
                                                    dataframe=train_data,
                                                    index="customer_id")
es_train_data

Next, we are ready to create our features, and feature matrix for the training data.

[ ]:
feature_matrix_train, feature_defs = ft.dfs(entityset=es_train_data,
                                            target_entity="customers")
feature_matrix_train

We will also encode our feature matrix to make machine learning compatible features.

[ ]:
feature_matrix_train_enc, features_enc = ft.encode_features(feature_matrix_train, feature_defs)
feature_matrix_train_enc.head()

Notice how the the whole feature matrix only inclues numeric values now.

Now we can use the feature definitions to calculate our feature matrix for the test data, and avoid label leakage.

2. Data with timestamps

If your data has timestamps, the best way to prevent label leakage is to use a list of cutoff times, which specify the last point in time data is allowed to be used for each row in the resulting feature matrix. To use cutoff times, you need to set a time index for each time sensitive entity in your entity set.

Tip: Even if your data doesn’t have time stamps, you could add a column with dummy timestamps that can be used by Featuretools as time index.

When you call ft.dfs, you can provide a Dataframe of cutoff times like this:

[ ]:
cutoff_times = pd.DataFrame({"customer_id": [1, 2, 3, 4, 5],
                             "time": pd.date_range('2014-01-01 01:41:50', periods=5, freq='25min')})
cutoff_times.head()
[ ]:
train_test_data = pd.DataFrame({"customer_id": [1, 2, 3, 4, 5],
                                "age": [20, 25, 55, 22, 35],
                                "gender": ["f", "m", "m", "m", "m"],
                                "signup_date": pd.date_range('2010-01-01 01:41:50', periods=5, freq='25min')})

es_train_test_data = ft.EntitySet(id="customer_train_test_data")
es_train_test_data = es_train_test_data.entity_from_dataframe(entity_id="customers",
                                                              dataframe=train_test_data,
                                                              index="customer_id",
                                                              time_index="signup_date")

feature_matrix_train_test, features = ft.dfs(entityset=es_train_test_data,
                                             target_entity="customers",
                                             cutoff_time=cutoff_times,
                                             cutoff_time_in_index=True)
feature_matrix_train_test.head()

Above, we have created a feature matrix that uses cutoff times to avoid label leakage. We could also encode this feature matrix using ft.encode_features.

What is the difference between passing a primitive object versus a string to DFS?

There are 2 ways to pass primitives to DFS: the primitive object, or a string of the primitive name.

We will use the Transform primitive called TimeSincePrevious to illustrate the differences.

First, let’s use the string of primitive name.

[ ]:
es = ft.demo.load_mock_customer(return_entityset=True)
[ ]:
feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="customers",
                                      agg_primitives=[],
                                      trans_primitives=["time_since_previous"])
feature_matrix

Now, let’s use the primitive object.

[ ]:
from featuretools.primitives import TimeSincePrevious

feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="customers",
                                      agg_primitives=[],
                                      trans_primitives=[TimeSincePrevious])
feature_matrix

As we can see above, the feature matrix is the same.

However, if we need to modify controllable parameters in the primitive, we should use the primitive object. For instance, let’s make TimeSincePrevious return units of hours (the default is in seconds).

[ ]:
from featuretools.primitives import TimeSincePrevious

time_since_previous_in_hours = TimeSincePrevious(unit='hours')

feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="customers",
                                      agg_primitives=[],
                                      trans_primitives=[time_since_previous_in_hours])
feature_matrix

Features

How can I select features based on some attributes (a specific string, an explicit primitive type, a return type, a given depth)?

You may wish to select a subset of your features based on some attributes.

Let’s say you wanted to select features that had the string amount in its name. You can check for this by using the get_name function on the feature definitions.

[ ]:
es = ft.demo.load_mock_customer(return_entityset=True)

feature_defs = ft.dfs(entityset=es,
                      target_entity="customers",
                      features_only=True)

features_with_amount = []
for x in feature_defs:
    if 'amount' in x.get_name():
        features_with_amount.append(x)
features_with_amount[0:5]

You might also want to only select features that are aggregation features.

[ ]:
from featuretools import AggregationFeature

features_only_aggregations = []
for x in feature_defs:
    if type(x) == AggregationFeature:
        features_only_aggregations.append(x)
features_only_aggregations[0:5]

Also, you might only want to select features that are calculated at a certain depth. You can do this by using the get_depth function.

[ ]:
features_only_depth_2 = []
for x in feature_defs:
    if x.get_depth() == 2:
        features_only_depth_2.append(x)
features_only_depth_2[0:5]

Finally, you might only want features that return a certain type. You can do this by using the variable_type function.

[ ]:
from featuretools.variable_types import Numeric

features_only_numeric = []
for x in feature_defs:
    if x.variable_type == Numeric:
        features_only_numeric.append(x)
features_only_numeric[0:5]

Once you have your specific feature list, you can use ft.calculate_feature_matrix to generate a feature matrix for only those features.

For our example, let’s use the features with only the string amount in its name.

[ ]:
feature_matrix = ft.calculate_feature_matrix(entityset=es,
                                             features=features_with_amount) # change to your specific feature list
feature_matrix.head()

Above, notice how all the column names for our feature matrix contain the string amount.

How do I create where features?

Sometimes, you might want to create features that are conditioned on a second value before it is calculated. This extra filter is called a “where clause”. You can create these features using the using the interesting_values of a variable.

If you have categorical columns in your EntitySet, you can use then add_interesting_values. This function will find interesting values for your categorical variables, which can then be used to generate “where” clauses.

First, let’s create our EntitySet.

[ ]:
es = ft.demo.load_mock_customer(return_entityset=True)
es

Now we can add the interesting variables for the categorical variables.

[ ]:
es.add_interesting_values()

Now we can run DFS with the where_primitives argument to define which primitives to apply with where clauses. In this case, let’s use the primitive count.

[ ]:
feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="customers",
                                      agg_primitives=["count"],
                                      where_primitives=["count"],
                                      trans_primitives=[])
feature_matrix.head()

We have now created some useful features. One example of a useful feature is the COUNT(sessions WHERE device = tablet). This feature tells us how many sessions a customer completed on a tablet.

[ ]:
feature_matrix[["COUNT(sessions WHERE device = tablet)"]]

Primitives

What is the difference between the primitive types (Transform, GroupBy Transform, & Aggregation)?

You might curious to know the difference between the primitive groups. Let’s review the differences between transform, groupby transform, and aggregation primitives.

First, let’s create a simple EntitySet.

[ ]:
import pandas as pd
import featuretools as ft

df = pd.DataFrame({
    "id": [1, 2, 3, 4, 5, 6],
    "time_index": pd.date_range("1/1/2019", periods=6, freq="D"),
    "group": ["a", "b", "a", "c", "a", "b"],
    "val": [5, 1, 10, 20, 6, 23],
})
es = ft.EntitySet()
es = es.entity_from_dataframe(entity_id="observations",
                              dataframe=df,
                              index="id",
                              time_index="time_index")

es = es.normalize_entity(base_entity_id="observations",
                         new_entity_id="groups",
                         index="group")

es.plot()

After calling normalize_entity, the variable “group” has the type “id” because it identifies another entity. Alternatively, it could be set using the variable_types parameter when we first call es.entity_from_dataframe().

Transform Primitive

The cum_sum primitive calculates the running sum in list of numbers.

[ ]:
from featuretools.primitives import CumSum

cum_sum = CumSum()
cum_sum([1, 2, 3, 4, 5]).tolist()

If we apply it using the trans_primitives argument it will calculate it over the entire observations entity like this:

[ ]:
feature_matrix, feature_defs = ft.dfs(target_entity="observations",
                                      entityset=es,
                                      agg_primitives=[],
                                      trans_primitives=["cum_sum"],
                                      groupby_trans_primitives=[])

feature_matrix

Groupby Transform Primitive

If we apply it using groupby_trans_primitives, then DFS will first group by any id variables before applying the transform primitive. As a result, we get the cumulative sum by group.

[ ]:
feature_matrix, feature_defs = ft.dfs(target_entity="observations",
                                      entityset=es,
                                      agg_primitives=[],
                                      trans_primitives=[],
                                      groupby_trans_primitives=["cum_sum"])

feature_matrix

Aggregation Primitive

Finally, there is also the aggregation primitive “sum”. If we use sum, it will calculate the sum for the group at the cutoff time for each row. Because we didn’t specify a cutoff time it will use all the data for each group for each row.

[ ]:
feature_matrix, feature_defs = ft.dfs(target_entity="observations",
                                      entityset=es,
                                      agg_primitives=["sum"],
                                      trans_primitives=[],
                                      cutoff_time_in_index=True,
                                      groupby_trans_primitives=[])

feature_matrix

If we set the cutoff time of each row to be the time index, then use sum as an aggregation primitive, the result is the same as cum_sum. (Though the order is different in the displayed dataframe).

[ ]:
cutoff_time = df[["id", "time_index"]]
cutoff_time
[ ]:
feature_matrix, feature_defs = ft.dfs(target_entity="observations",
                                      entityset=es,
                                      agg_primitives=["sum"],
                                      trans_primitives=[],
                                      groupby_trans_primitives=[],
                                      cutoff_time_in_index=True,
                                      cutoff_time=cutoff_time)

feature_matrix

How do I get a list of all Aggregation and Transform primitives?

You can do featuretools.list_primitives() to get all the primitive in Featuretools. It will return a Dataframe with the names, type, and description of the primitives.

[ ]:
df_primitives = ft.list_primitives()
df_primitives.head()
[ ]:
df_primitives.tail()

How do I change the units for a TimeSince primitive?

There are a few primitives in Featuretools that make some time-based calculation. These include TimeSince, TimeSincePrevious, TimeSinceLast, TimeSinceFirst.

You can change the units from the default seconds to any valid time unit, by doing the following:

[ ]:
from featuretools.primitives import TimeSince, TimeSincePrevious, TimeSinceLast, TimeSinceFirst

time_since = TimeSince(unit="minutes")
time_since_previous = TimeSincePrevious(unit="hours")
time_since_last = TimeSinceLast(unit="days")
time_since_first = TimeSinceFirst(unit="years")

es = ft.demo.load_mock_customer(return_entityset=True)

feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="customers",
                                      agg_primitives=[time_since_last, time_since_first],
                                      trans_primitives=[time_since, time_since_previous])

Above, we changed the units to the following: - minutes for TimeSince - hours for TimeSincePrevious - days for TimeSinceLast - years for TimeSinceFirst.

Now we can see that our feature matrix contains multiple features where the units for the TimeSince primitives are changed.

[ ]:
feature_matrix.head()

There are now features where time unit is different from the default of seconds, such as TIME_SINCE_LAST(sessions.session_start, unit=days), and TIME_SINCE_FIRST(sessions.session_start, unit=years).

Modeling

How does my train & test data work with Featuretools and sklearn’s train_test_split?

You might be wondering how to properly use your train & test data with Featuretools, and sklearn’s train_test_split. There are a few things you must do to ensure accuracy with this workflow.

Let’s imagine we have a Dataframes for our train data, with the labels.

[ ]:
train_data = pd.DataFrame({"customer_id": [1, 2, 3, 4, 5],
                           "age": [20, 25, 55, 22, 35],
                           "gender": ["f", "m", "m", "m", "m"],
                           "signup_date": pd.date_range('2010-01-01 01:41:50', periods=5, freq='25min'),
                           "labels": [False, True, True, False, False]})
train_data.head()

Now we can create our EntitySet for the train data, and create our features. To prevent label leakage, we will use cutoff times (see earlier question).

[ ]:
es_train_data = ft.EntitySet(id="customer_data")
es_train_data = es_train_data.entity_from_dataframe(entity_id="customers",
                                                    dataframe=train_data,
                                                    index="customer_id")

cutoff_times = pd.DataFrame({"customer_id": [1, 2, 3, 4, 5],
                             "time": pd.date_range('2014-01-01 01:41:50', periods=5, freq='25min')})

feature_matrix_train, features = ft.dfs(entityset=es_train_data,
                                        target_entity="customers",
                                        cutoff_time=cutoff_times,
                                        cutoff_time_in_index=True)
feature_matrix_train.head()

We will also encode our feature matrix to compatible for machine learning algorithms.

[ ]:
feature_matrix_train_enc, feature_enc = ft.encode_features(feature_matrix_train, features)
feature_matrix_train_enc.head()
[ ]:
from sklearn.model_selection import train_test_split

X = feature_matrix_train_enc.drop(['labels'], axis=1)
y = feature_matrix_train_enc['labels']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25)

Now you can use the encoded feature matrix with sklearn’s train_test_split. This will allow you to train your model, and tune your parameters.

How are categorical variables encoded when splitting training and testing data?

You might be wondering what happens when categorical variables are encoded with your training and testing data. You might be curious to know what happens if the train data has a categorical variable that is not present in the testing data.

Let’s explore a simple example to see what happens during the encoding process.

[ ]:
train_data = pd.DataFrame({"customer_id": [1, 2, 3, 4, 5],
                           "product_purchased": ["coke zero", "car", "toothpaste", "coke zero", "car"]})

es_train = ft.EntitySet(id="customer_data")
es_train = es_train.entity_from_dataframe(entity_id="customers",
                                          dataframe=train_data,
                                          index="customer_id")

feature_matrix_train, features = ft.dfs(entityset=es_train,
                                        target_entity='customers')
feature_matrix_train

We will use ft.encode_features to properly encode the product_purchased column.

[ ]:
feature_matrix_train_encoded, features_encoded = ft.encode_features(feature_matrix_train,
                                                                    features)
feature_matrix_train_encoded.head()

Now lets imagine we have some test data that has doesn’t have one of the categorical values (toothpaste). Also, the test data has a value that wasn’t present in the train data (water).

[ ]:
test_data = pd.DataFrame({"customer_id": [6, 7, 8, 9, 10],
                          "product_purchased": ["coke zero", "car", "coke zero", "coke zero", "water"]})

es_test = ft.EntitySet(id="customer_data")
es_test = es_test.entity_from_dataframe(entity_id="customers",
                                        dataframe=test_data,
                                        index="customer_id")

feature_matrix_test = ft.calculate_feature_matrix(entityset=es_test,
                                                  features=features_encoded)
feature_matrix_test.head()

As seen above, we were able to successfully handle the encoding, and deal with the following complications: - toothpaste was present in the training data but not present in the testing data - water was present in the test data but not present in the training data.

Errors & Warnings

Why am I getting this error ‘Index is not unique on dataframe’?

You may be trying to create your EntitySet, and run into this error.

AssertionError: Index is not unique on dataframe

This is because each entity in your EntitySet needs a unique index.

Let’s look at a simple example.

[ ]:
product_df = pd.DataFrame({'id': [1, 2, 3, 4, 4],
                           'rating': [3.5, 4.0, 4.5, 1.5, 5.0]})
product_df

Notice how the id column has a duplicate index of 4. If you try to create an entity with this Dataframe, you will run into the following error.

es = ft.EntitySet(id="product_data")
es = es.entity_from_dataframe(entity_id="products",
                              dataframe=product_df,
                              index="id")
---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
<ipython-input-63-a6e02ba6fa47> in <module>
      2 es = es.entity_from_dataframe(entity_id="products",
      3                               dataframe=product_df,
----> 4                               index="id")

~/featuretools/featuretools/entityset/entityset.py in entity_from_dataframe(self, entity_id, dataframe, index, variable_types, make_index, time_index, secondary_time_index, already_sorted)
    486             secondary_time_index=secondary_time_index,
    487             already_sorted=already_sorted,
--> 488             make_index=make_index)
    489         self.entity_dict[entity.id] = entity
    490         self.reset_data_description()

~/featuretools/featuretools/entityset/entity.py in __init__(self, id, df, entityset, variable_types, index, time_index, secondary_time_index, last_time_index, already_sorted, make_index, verbose)
     79
     80         self.df = df[[v.id for v in self.variables]]
---> 81         self.set_index(index)
     82
     83         self.time_index = None

~/featuretools/featuretools/entityset/entity.py in set_index(self, variable_id, unique)
    450         self.df.index.name = None
    451         if unique:
--> 452             assert self.df.index.is_unique, "Index is not unique on dataframe (Entity {})".format(self.id)
    453
    454         self.convert_variable_type(variable_id, vtypes.Index, convert_data=False)

AssertionError: Index is not unique on dataframe (Entity products)

To fix the above error, you can do one of the following solutions:

Solution #1 - You can create a unique index on your Dataframe.

[ ]:
product_df = pd.DataFrame({'id': [1, 2, 3, 4, 5],
                           'rating': [3.5, 4.0, 4.5, 1.5, 5.0]})
product_df

Notice how we now have a unique index column called id.

[ ]:
es = es.entity_from_dataframe(entity_id="products",
                              dataframe=product_df,
                              index="id")
es

As seen above, we can now create our entity for our EntitySet without an error by creating a unique index in our Dataframe.

Solution #2 - Set make_index to True in your call to entity_from_dataframe to create a new index on that data - make_index creates a unique index for each row by just looking at what number the row is, in relation to all the other rows.

[ ]:
product_df = pd.DataFrame({'id': [1, 2, 3, 4, 4],
                           'rating': [3.5, 4.0, 4.5, 1.5, 5.0]})

es = ft.EntitySet(id="product_data")
es = es.entity_from_dataframe(entity_id="products",
                              dataframe=product_df,
                              index="product_id",
                              make_index=True)
es['products'].df

As seen above, we created our entity for our EntitySet without an error using the make_index argument.

Why am I getting the following warning ‘Using training_window but last_time_index is not set’?

If you are using a training window, and you haven’t set a last_time_index for your entity, you will get this warning. The training window attribute in Featuretools limits the amount of past data that can be used while calculating a particular feature vector.

You can add the last_time_index to all entities automatically by calling your_entityset.add_last_time_indexes() after you create your EntitySet. This will remove the warning.

[ ]:
es = ft.demo.load_mock_customer(return_entityset=True)
es.add_last_time_indexes()

Now we can run DFS without getting the warning.

[ ]:
cutoff_times = pd.DataFrame()
cutoff_times['customer_id'] = [1, 2, 3, 1]
cutoff_times['time'] = pd.to_datetime(['2014-1-1 04:00', '2014-1-1 05:00', '2014-1-1 06:00', '2014-1-1 08:00'])
cutoff_times['label'] = [True, True, False, True]

feature_matrix, feature_defs = ft.dfs(entityset=es,
                                      target_entity="customers",
                                      cutoff_time=cutoff_times,
                                      cutoff_time_in_index=True,
                                      training_window="1 hour")

last_time_index vs. time_index

  • The time_index is when the instance was first known.

  • The last_time_index is when the instance appears for the last time.

  • For example, a customer’s session has multiple transactions which can happen at different points in time. If we are trying to count the number of sessions a user has in a given time period, we often want to count all the sessions that had any transaction during the training window. To accomplish this, we need to not only know when a session starts (time_index), but also when it ends (last_time_index). The last time that an instance appears in the data is stored as the last_time_index of an Entity.

  • Once the last_time_index has been set, Featuretools will check to see if the last_time_index is after the start of the training window. That, combined with the cutoff time, allows DFS to discover which data is relevant for a given training window.

Why am I getting errors with Featuretools on Google Colab?

Google Colab, by default, has Featuretools 0.4.1 installed. You may run into issues following our newest guides, or latest documentation while using an older version of Featuretools. Therefore, we suggest you upgrade to the latest featuretools version by doing the following in your notebook in Google Colab:

!pip install -U featuretools

You may need to Restart the runtime by doing Runtime -> Restart Runtime. You can check latest Featuretools version by doing following:

import featuretools as ft
print(ft.__version__)

You should see a version greater than 0.4.1