Pairs of restaurants

In the last lesson, you cleaned the restaurants dataset to make it ready for building a restaurants recommendation engine. You have a new DataFrame named restaurants_new with new restaurants to train your model on, that's been scraped from a new data source.

You've already cleaned the cuisine_type and city columns using the techniques learned throughout the course. However you saw duplicates with typos in restaurants names that require record linkage instead of joins with restaurants.

In this exercise, you will perform the first step in record linkage and generate possible pairs of rows between restaurants and restaurants_new. Both DataFrames, pandas and recordlinkage are in your environment.

This exercise is part of the course

Cleaning Data in Python

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# Create an indexer and object and find possible pairs
indexer = ____

# Block pairing on cuisine_type
indexer.____(____)

# Generate pairs
pairs = indexer.____(____, ____)

Edit and Run Code

This exercise is part of the course

Cleaning Data in Python

IntermediateSkill Level

4.8+

Start Course for Free

In this chapter, you'll learn how to overcome some of the most common dirty data problems. You'll convert data types, apply range constraints to remove future data points, and remove duplicated data points to avoid double-counting.

Exercise 1: Data type constraints Exercise 2: Common data types Exercise 3: Numeric data or ... ?Exercise 4: Summing strings and concatenating numbers Exercise 5: Data range constraints Exercise 6: Tire size constraints Exercise 7: Back to the future Exercise 8: Uniqueness constraints Exercise 9: How big is your subset?Exercise 10: Finding duplicates Exercise 11: Treating duplicates

Categorical and text data can often be some of the messiest parts of a dataset due to their unstructured nature. In this chapter, you’ll learn how to fix whitespace and capitalization inconsistencies in category labels, collapse multiple categories into one, and reformat strings for consistency.

Exercise 1: Membership constraints Exercise 2: Members only Exercise 3: Finding consistency Exercise 4: Categorical variables Exercise 5: Categories of errors Exercise 6: Inconsistent categories Exercise 7: Remapping categories Exercise 8: Cleaning text data Exercise 9: Removing titles and taking names Exercise 10: Keeping it descriptive

In this chapter, you’ll dive into more advanced data cleaning problems, such as ensuring that weights are all written in kilograms instead of pounds. You’ll also gain invaluable skills that will help you verify that values have been added correctly and that missing values don’t negatively impact your analyses.

Exercise 1: Uniformity Exercise 2: Ambiguous dates Exercise 3: Uniform currencies Exercise 4: Uniform dates Exercise 5: Cross field validation Exercise 6: Cross field or no cross field?Exercise 7: How's our data integrity?Exercise 8: Completeness Exercise 9: Is this missing at random?Exercise 10: Missing investors Exercise 11: Follow the money

Record linkage is a powerful technique used to merge multiple datasets together, used when values have typos or different spellings. In this chapter, you'll learn how to link records by calculating the similarity between strings—you’ll then use your new skills to join two restaurant review datasets into one clean master dataset.

Exercise 1: Comparing strings Exercise 2: Minimum edit distance Exercise 3: The cutoff point Exercise 4: Remapping categories II Exercise 5: Generating pairs Exercise 6: To link or not to link?Exercise 7: Pairs of restaurants

Current Exercise

Exercise 8: Similar restaurants Exercise 9: Linking DataFrames Exercise 10: Getting the right index Exercise 11: Linking them together!Exercise 12: Congratulations!