Aggregation-based features in Data Analysis Python - Cheat Sheet & Quick Revision

Recall & Review

beginner

What are aggregation-based features in data science?

Aggregation-based features are new data columns created by summarizing or combining existing data points, like calculating averages, sums, or counts over groups of data.

beginner

Why do we use aggregation-based features in machine learning?

We use aggregation-based features to capture group-level patterns and trends that single data points might miss, helping models learn better from the data.

beginner

Example: What does the aggregation 'mean' represent when applied to a group of data?

The 'mean' is the average value of the data points in the group, showing the central tendency of that group.

intermediate

How can you create aggregation-based features using Python's pandas library?

You can use pandas' groupby() function to group data by one or more columns, then apply aggregation functions like mean(), sum(), count(), etc., to create new features.

intermediate

What is the difference between 'count' and 'nunique' aggregations?

'Count' gives the total number of non-null data points in a group, while 'nunique' counts how many unique values are in that group.

Which of the following is NOT an aggregation function?

A filter

B sum

C count

D mean

Filter is used to select data, not to aggregate or summarize it.

What does the 'groupby' function in pandas do?

A Sorts the data

B Splits data into groups based on column values

C Deletes duplicate rows

D Creates new columns

Groupby splits data into groups so you can apply aggregation functions on each group.

If you want to find the total sales per store, which aggregation would you use?

A mean

B nunique

C count

D sum

Sum adds all sales values to get the total per store.

What does the 'nunique' aggregation tell you?

A Average value

B Total sum of values

C Number of unique values in a group

D Maximum value

Nunique counts how many different unique values exist in the group.

Why might aggregation-based features improve a model's performance?

A They capture group-level patterns

B They add noise to the data

C They reduce the number of rows

D They remove missing values

Aggregation-based features help models learn from group trends and patterns.

Explain how you would create an aggregation-based feature to find the average purchase amount per customer using pandas.

Think about grouping data by customer and then calculating the average.

Describe the difference between 'count' and 'nunique' aggregations and give an example of when each might be useful.

Consider counting total purchases vs. counting unique products bought.

Practice

1. What is the main purpose of aggregation-based features in data analysis?

easy

A. To summarize data by groups using functions like sum or mean

B. To visualize data using charts and graphs

C. To clean missing values from the dataset

D. To split data into training and testing sets

Solution

1. Step 1: Understand aggregation-based features

   Aggregation-based features summarize data by groups using functions like sum, mean, or count.

2. Step 2: Compare with other options

   Visualization, cleaning, and splitting data are different tasks not related to aggregation features.

3. Final Answer:

   To summarize data by groups using functions like sum or mean -> Option A

4. Quick Check:

   Aggregation = summary by groups [OK]

Hint: Aggregation means grouping and summarizing data [OK]

Common Mistakes:

• Confusing aggregation with visualization
• Thinking aggregation cleans data
• Mixing aggregation with data splitting

2. Which of the following is the correct syntax to calculate the mean of column 'sales' grouped by 'region' in pandas?

easy

A. df.mean('sales').groupby('region')

B. df['sales'].groupby('region').mean()

C. df.groupby('region')['sales'].mean()

D. df.groupby('sales')['region'].mean()

Solution

1. Step 1: Identify correct groupby syntax

   The correct syntax groups by 'region' then selects 'sales' column and applies mean: df.groupby('region')['sales'].mean()

2. Step 2: Check incorrect options

   df.mean('sales').groupby('region') reverses order, df['sales'].groupby('region').mean() tries to group on a column series, df.groupby('sales')['region'].mean() groups by 'sales' instead of 'region'.

3. Final Answer:

   df.groupby('region')['sales'].mean() -> Option C

4. Quick Check:

   Group by region, then mean sales [OK]

Hint: Use df.groupby('col')['col2'].agg_func() [OK]

Common Mistakes:

• Calling mean before groupby
• Grouping on wrong column
• Applying groupby on a Series instead of DataFrame

3. Given the DataFrame:

import pandas as pd
data = {'team': ['A', 'A', 'B', 'B', 'B'], 'points': [10, 15, 10, 5, 20]}
df = pd.DataFrame(data)
result = df.groupby('team').agg({'points': 'sum'})
print(result)

What is the output?

medium

A. team\nA 25\nB 10

B. team\nA 10\nB 20

C. team\nA 15\nB 10

D. team\nA 25\nB 35

Solution

1. Step 1: Group data by 'team'

   Teams are 'A' and 'B'. 'A' has points 10 and 15, 'B' has 10, 5, and 20.

2. Step 2: Sum points per team

   Sum for 'A' = 10 + 15 = 25; Sum for 'B' = 10 + 5 + 20 = 35.

3. Final Answer:

   team A 25 B 35 -> Option D

4. Quick Check:

   Sum points by team = 25 and 35 [OK]

Hint: Sum points after groupby to get totals per team [OK]

Common Mistakes:

• Adding only first value per group
• Confusing sum with mean
• Misreading group labels

4. What is wrong with this code snippet to get the average 'score' per 'class' in pandas?

df.groupby('class').score.agg('average')

medium

A. The groupby column 'class' is misspelled

B. The aggregation function name 'average' is invalid; should be 'mean'

C. The code should use double brackets: df.groupby(['class'])

D. The code is correct and will run without error

Solution

1. Step 1: Check aggregation function name

   'average' is not a valid pandas aggregation function; the correct function for average is 'mean'.

2. Step 2: Verify other parts

   Grouping by 'class' and accessing 'score' column is correct syntax; double brackets are optional here.

3. Final Answer:

   The aggregation function name 'average' is invalid; should be 'mean' -> Option B

4. Quick Check:

   Use 'mean' not 'average' for average calculation [OK]

Hint: Use 'mean' for average in agg() [OK]

Common Mistakes:

• Using 'average' instead of 'mean'
• Thinking double brackets are mandatory
• Assuming code runs without error

5. You have a sales DataFrame with columns 'store', 'product', and 'revenue'. You want to create a new feature showing the average revenue per store and product combined. Which pandas code correctly creates this aggregation-based feature and adds it as a new column 'avg_rev' to the original DataFrame?

hard

A. df['avg_rev'] = df.groupby(['store', 'product'])['revenue'].transform('mean')

B. df['avg_rev'] = df.groupby('store')['revenue'].mean()

C. df['avg_rev'] = df.groupby('product')['revenue'].sum()

D. df['avg_rev'] = df.groupby(['store', 'product']).agg({'revenue': 'mean'})

Solution

1. Step 1: Understand the goal

   We want average revenue per store and product, and add it as a new column matching original rows.

2. Step 2: Use groupby with transform

   Using groupby on ['store', 'product'] and transform('mean') returns a Series aligned with original DataFrame, perfect for new column.

3. Step 3: Check other options

   df['avg_rev'] = df.groupby('store')['revenue'].mean() groups only by store, not product; df['avg_rev'] = df.groupby('product')['revenue'].sum() sums revenue, not average; df['avg_rev'] = df.groupby(['store', 'product']).agg({'revenue': 'mean'}) returns a DataFrame, not a Series aligned for assignment.

4. Final Answer:

   df['avg_rev'] = df.groupby(['store', 'product'])['revenue'].transform('mean') -> Option A

5. Quick Check:

   Use transform to add group average as new column [OK]

Hint: Use transform for group stats as new column [OK]

Common Mistakes:

• Using agg() which returns DataFrame, not Series
• Grouping by only one column instead of both
• Using sum instead of mean