Join our book community on Discord

https://packt.link/EarlyAccessCommunity

We continue our journey around the world of data by exploring two datasets with geographical distributed data. The first dataset is Every Pub in England. This dataset contains the unique id, name, address, postcode and geographical position data for every pub in England. The second dataset is Starbucks Locations Worldwide. This dataset contains the store number, name, ownership, as well as street address, city and geographical information (latitude and longitude) for all Starbucks stores in the world. We will combine these two datasets and will add also additional geographical support data. We will learn how to work with missing data and perform data imputation if needed, how to visualize geographical data, how to clip and merge polygon data, how to generate custom maps and create multiple layers over maps.

Pubs in England

The dataset contains data about 51,566 pubs in England, including the pub name, the address, the postal code, the geographical position (both by easting and northing and by latitude and longitude) and the local authority. I created a Notebook, Every Pub in England – Data Exploration to investigate this data.

Data quality check

For the data quality check, we will use info() and describe() to get a first glimpse. Then, we can also use our custom data quality statistics functions. We saw in the previous chapter these functions, will not repeat here. Because we will keep using them, we will group them in a utility script. I called this utility script data_quality_stats and I defined in this module the functions missing_data, most_frequent_values and unique_values. To use the functions defined in this utility script, we need to first add it to the Notebook. From File menu, we select Add utility script menu item. Then, we add the import in one of the first Notebook cells...

Starbucks in the World

We start the analysis of Starbucks Locations Worldwide dataset with a detailed exploratory data analysis (EDA) in the notebook Starbucks Location Worldwide - Data Exploration. The tools used in this dataset are imported from data_quality_stats and from plot_style_utils utility scripts. Before starting our analysis, it is important to explain that the dataset used for this analysis is from Kaggle and was collected 6 years ago. Meantime, Starbucks business expanded very much and therefore the number of shops, the geographical distribution of the shops, all this information is not up to date.

Preliminary data analysis

The dataset has 25,600 rows, with only 1 latitude and longitude values missing, 2 Street Addresses, 15 Cities. The fields that have the most missing data are Postcode (5.9%) and Phone Number (26.8%). In Figure 3.16 we can see few a sample of the data.

Figure 4.16. First rows of Starbucks dataset

Looking to the most frequent values report, we can learn...

Pubs and Starbucks in London

Until now our analysis was focused on individual datasets Every Pub in England and Starbucks Locations Worldwide. To support some of the data analysis tasks related to these two separate datasets, we also added two more datasets, one with geographical position of postal codes, to replace missing latitude and longitude data and one shapefile data for United Kingdom, used to clip the Voronoi polygons generated from pubs position, to align them to the land contour of the islands.In the following we will combine the information from the two main data sources analysed separately and will apply methods developed during this preliminary analysis to support the objective of our study. This will focus on a smaller region, where we have both a high density of pubs and also a concentration of Starbucks coffee shops, London. We already can hypothesize that the geospatial concentration of Starbucks is smaller than the concentration of pubs. We would like to see what the...

Summary

In this chapter we learned how to work with geographical information and maps, how to manipulate geometry data (clip and merge polygons data, cluster data to generate maps with less details, remove subsets of geospatial data), superpose several layers of data over maps. We also learned how to modify and extract information from shapefile using geopandas and custom code as well as creating or calculating geospatial features, like terrain area or geospatial objects density. Additionally, we extracted reusable functions and grouped them in two utility scripts, which is Kaggle wording for independent Python modules. These utility scripts can be imported as any other library and integrated with your Notebook code. In the next Chapter we will put at work some of these tools and techniques for a data analytics competition.

References

1. Every Pub in England, Kaggle Datasets, https://www.kaggle.com/datasets/rtatman/every-pub-in-england
2. Starbucks Locations Worldwide, Kaggle Datasets, https...

References

1. Data Science for Good: Kiva Crowdfunding, Kaggle dataset: https://www.kaggle.com/kiva/data-science-for-good-kiva-crowdfunding
2. Meta Kaggle, Kaggle dataset: https://www.kaggle.com/datasets/kaggle/meta-kaggle
3. Country Statistics – UNData, Kaggle dataset: https://www.kaggle.com/sudalairajkumar/undata-country-profiles
4. Kiva Microloans – A Data Exploration, Kaggle notebook: https://github.com/PacktPublishing/Developing-Kaggle-Notebooks/blob/develop/Chapter-05/kiva-microloans-a-data-exploration.ipynb
5. Multidimensional poverty index (MPI): http://hdr.undp.org/en/content/multidimensional-poverty-index-mpi
6. Multidimensional poverty index on Wikipedia: https://en.wikipedia.org/wiki/Multidimensional_Poverty_Index
7. plotly-utils, Kaggle utility script: https://github.com/PacktPublishing/Developing-Kaggle-Notebooks/blob/develop/Chapter-05/plotly-utils.ipynb
8. Kiva: Loans that change lives: https://theglobalheroes.wordpress.com/2012...