Capstone Project

Robert Joseph

July 18, 2020

 Abstract

This report discusses the use of Foursquare Api to build upon the
comparison between Paris and London and the correlation whether
someone should start a buisness in either of the two cities. Various
machine learning algorithms were used to compare and differentiate
the two cities as well as other geolocation tools to check either of
the city is good to start an Artificial Intelligence Business.
1 - Introduction
The final course of the Data Science Professional Certificate consist
of a capstone project where in all the skills and relevant knowledge
that one has gathered from this 9 intense courses has to be applied
on a final capstone project.

The final problem as well as the analysis is the left for the reader to
explore and decide. The idea uses location data with the help of the
foursquare api that can be leveraged into coming up with a problem
that the foursquare location data to solve it or just in contrast to
compare cities or neighbourhoods of ones own choice.

London is the capital and largest city of England and the United
Kingdom.Standing on the River Thames in the south-east of Eng-
land, at the head of its 50-mile (80 km) estuary leading to the North
Sea, London has been a major settlement for two millennia.

Paris is the capital and most populous city of France, with an es-
timated population of 2,150,271 residents as of 2020, in an area of
105 square kilometres (41 square miles).Since the 17th century, Paris
has been one of Europe’s major centres of finance, diplomacy, com-
merce, fashion, science and arts.

The main Goal of this project that I have chosen would be to eval-
uate the comparison between Paris and London as well as point out
the differences. Another factor to be included is which city would
be more ideal to start an Artificial Intelligence company and the
various factors correlating to it as both cities are major cities and
global hotspots in the world for tech companies.

1
Target Audience

ˆ Potential Entreprenuers who want to start a business relating

to Machine learning /AI/data science.
ˆ People who wanna choose which city to live in the future
ˆ Course instructors and learners who will grade this project as
well as showcase what I have learnt through this course.

2
2 - Business Problem
In this ever changing world of technology and reforms the use of
AI will dominate and change most of the world and industries as we
know so among the two busiest cities in the world which one would
a person be willing to start a business in AI. Various factors would
be included such as pricing, multiculturism, language barriers and
so on would influence this decision.

3
 Figure 1: Paris Geolocation Dataset

3 - Data
Various data sets were collected, reformatted and analysed in
order to get the required results. Some of them include
ˆ http://www.cgedd.developpement-durable.gouv.fr/house-prices-
in-france-property-price-index-french-a1117.html - House Prices
in France
ˆ https://www.kaggle.com/alphaepsilon/housing-prices-dataset -
Housing Dataset
ˆ https://data.world/datasets/real-estate - Numerous Datasets
for different categories
ˆ https://data.london.gov.uk/dataset?tag=start-ups - Data sets
for london
ˆ https://www.kaggle.com/tags/companies - Various companies
and their datasets
More datasets were included and merged to get the final dataset re-
lating to the idea.The use of even foursquare datasets were used and
important features such as Housing Prices, Locality, Famous icons,
Restaurant Prices, transportation facilities, technological hotspots
as well as access to a high wifi speed and so on were all assessed.

London

Almost half a million lines of records were present for the London
dataset and had to sampled such that only 160 rows were extracted

4
 Figure 2: London Geolocation Dataset

as it had too many records to handle.

This was achievable with the help of
# Python Code
rlondon = london.sample(frac = 0.0005) # only 0.0005 % of the data was
randomly selected
rlondon.head()

Various columns which served no purpose to the analysis were dropped

such as ’Police Station Code’, Lower layer super output area’, ’Ru-
ral/urban’,’Region’, ’Altitude’, ’London zone’ etc.
Finally the resulting dataset is shown in Figure 2

Paris

Similar to the London Dataset there were about 30,000 rows out
of which only a sample was taken by using the same python code as
state above.
The paris dataset had no columns which needed to be dropped and
so was retained in its original state.

Artificial Intelligence

In regard to analysing which city would be best suited for a new AI

startup a handful of datasets were extracted using web scarpping
tools.The final dataset was then merged and only the companies
that were located in London and Paris were extracted. The code
below shows how it was achieved.

5
import requests
import pandas as pd

url = ’https://golden.com/list-of-artificial-intelligence-companies/’
html = requests.get(url).content
df_list = pd.read_html(html)
df = df_list[-1]
print(df)
df.to_csv(’my data.csv’)

These are some of the links

ˆ http://analytics.dkv.global/data/pdf/AI-in-UK/AI-in-UK-1000-
UK-AI-Companies-Profiles.pdf
ˆ https://craft.co/artificial-intelligence-companies-in-paris-fr?page=1
ˆ https://clutch.co/fr/developers/artificial-intelligence
Finally to compare them various visualisation tools were used and
articles referenced in order to reach the final conclusion.

6
4 - Methodology
An in-depth research of the dataset has been done and a thorough
analysis of the various features and methods have been investigated
to ensure the maximum accuracy of the model as possible.

After reduction of the number of features in the data frame by

replacing them with more useful data cluster analysis was done to
find the best cluster of both Paris and London and then correlation
and various other visual graphs were used to compare the two cities.

GeoLocation

The algorithm below gets the required latitude and longitude of

London(similar code has been coded for Paris) using the Google
Maps Geocoder API.
address = "London, UK"

geolocator = Nominatim(user_agent="uk_explorer")
location = geolocator.geocode(address)
latitude = location.latitude
longitude = location.longitude
print(’The geographical coordinates of London are {},
{}.’.format(latitude, longitude))

Folium

Folium makes it easy to visualize data that’s been manipulated

in Python on an interactive leaflet map. It enables both the binding
of data to a map for choropleth visualizations as well as passing rich
vector/raster/HTML visualizations as markers on the map. It uses
the Open StreetMap technology.
The code below shows only of Paris but a similar code has been
coded even for London.

7
 Figure 3: Map of Paris with Markers

# Python Code
# create map of Paris using latitude and longitude values
map_paris = folium.Map(location = [latitude, longitude], zoom_start = 11)

# add markers to map

for lat, lng, county, name in zip(rparis[’Latitude’],
rparis[’Longitude’], rparis[’County’], rparis[’Place Name’]):
label = ’{}, {}’.format(county, name)
label = folium.Popup(label, parse_html = True)
folium.CircleMarker(
[lat, lng],
radius = 5,
popup = label,
color = ’red’,
fill = True,
fill_color = ’#3186cc’,
fill_opacity = 0.7,
parse_html = False).add_to(map_paris)

map_paris # show the map of paris with markers from the dataset

8
 Figure 4: Map of London with Markers

Foursquare API

The Foursquare API allows application developers to interact with

the Foursquare platform. With the help of the Foursquare API
venues and various other location and landmarks were extracted
and merged into a dataframe.
LIMIT = 10 # limit of number of venues returned by Foursquare API

radius = 500 # define radius

url =
’https://api.foursquare.com/v2/venues/explore?&client_id={}&client_secret={}&v={}&ll={},{}&radius=
CLIENT_ID,
CLIENT_SECRET,
VERSION,
neighborhood_latitude,
neighborhood_longitude,
radius,
LIMIT)
results = requests.get(url).json()

# function that extracts the category of the venue

def get_category_type(row):
try:

9
categories_list = row[’categories’]
except:
categories_list = row[’venue.categories’]

if len(categories_list) == 0:
return None
else:
return categories_list[0][’name’]

venues = results[’response’][’groups’][0][’items’]

nearby_venues = json_normalize(venues) # flatten JSON

# filter columns
filtered_columns = [’venue.name’, ’venue.categories’,
’venue.location.lat’, ’venue.location.lng’]
nearby_venues = nearby_venues.loc[:, filtered_columns]

# filter the category for each row

nearby_venues[’venue.categories’] =
nearby_venues.apply(get_category_type, axis = 1)

# clean columns
nearby_venues.columns = [col.split(".")[-1] for col in
nearby_venues.columns]

nearby_venues.head()

One Hot Encoding

One hot encoding is a process by which categorical variables are

converted into a form that could be provided to ML algorithms to
do a better job in prediction.
# one hot encoding
paris_onehot = pd.get_dummies(paris_venues[[’Venue Category’]], prefix =
"", prefix_sep = "")

# add neighborhood column back to dataframe

paris_onehot[’Neighborhood’] = paris_venues[’Neighborhood’]

# move neighborhood column to the first column

fixed_columns = [paris_onehot.columns[-1]] +
list(paris_onehot.columns[:-1])
paris_onehot = paris_onehot[fixed_columns]

paris_onehot.head()

10
 Figure 5: Top 10 Most visited Venues for London

Figure 6: Top 10 Most visited Venues for Paris

Most Visited Venues

The top 10 most visited venues were extracted from each neigh-
bourhood and then merged together to form another dataset.
def return_most_common_venues(row, num_top_venues):
row_categories = row.iloc[1:]
row_categories_sorted = row_categories.sort_values(ascending = False)
return row_categories_sorted.index.values[0:num_top_venues]

num_top_venues = 10

indicators = [’st’, ’nd’, ’rd’]

# create columns according to number of top venues

columns = [’Neighborhood’]
for ind in np.arange(num_top_venues):
try:
columns.append(’{}{} Most Common Venue’.format(ind+1,
indicators[ind]))
except:
columns.append(’{}th Most Common Venue’.format(ind+1))

# create a new dataframe

11
neighborhoods_venues_sorted = pd.DataFrame(columns=columns)
neighborhoods_venues_sorted[’Neighborhood’] =
paris_grouped[’Neighborhood’]

for ind in np.arange(paris_grouped.shape[0]):

neighborhoods_venues_sorted.iloc[ind, 1:] =
return_most_common_venues(paris_grouped.iloc[ind, :],
num_top_venues)

neighborhoods_venues_sorted.head()

K- Means Clustering

After the venues were put into a dataframe, The K- Means Cluster-
ing Machine Learning Algorithm was used to train the data and get
the desired clusters. The first task was finding the optimal K and
as there were two different datasets to explore
ˆ For Paris the optimal K found out to be was - 5
ˆ For London the optimal K found out to be was - 6
After finding the Optimal K the data was trained using KMeans
# set number of clusters
kclusters = int(len(rlondon["District"].unique()) / 4)
london_grouped_clustering = london_grouped.drop(’Neighborhood’, 1)

# run k-means clustering

kmeans = KMeans(n_clusters = kclusters, random_state =
1).fit(london_grouped_clustering)

Finally the data was then grouped into clusters as shown

12
 Figure 7: Map of Paris Clustered

Figure 8: Map of London Clustered

13
Figure 9: Countries with the maximum number of Artificial Intelligence Com-
panies/Startups

Artificial Intelligence

For the artificial intelligence dataset after cleaning and webscrap-

ping the number of startups were plotted and number of AI compa-
nies in each country was counted. Our main focus however is on the
United Kingdom specifically London and France specifically Paris.

14
5 - Results and Discussion
Finally we have reached the main part of our report. Let us break
this down into two parts

Comparison of London and Paris

ˆ Similarities
– Both cities are multicultural and diverse in their own ways
and share a rich history of their own.
– Most of the famous neighbourhoods have a restaurant as
its top most Common Venue.
– Example : In Paris the Louvre is one of the most famous
icons if not in the world also and its most commont venue
is the Plazza/French Restaurant.
– Similarly for the famous icon in London that is Westmin-
ister are Pubs and restaurants.
– The top 3 most common Venue points for London are
* Coffee Shop
* Hotel
* Cafe
– The top 3 most common Venue points for Paris are
* Coffee Shop
* Pub
* Cafe
– Both have an almost comparable population size of about
8-10 million.
– Both have an overwhelming power of attraction.
– When comparing the prices of the venues both of them are
expensive in their own ways and offer high quality food,
concerts, exhibition etc.

15
 ˆ Differences
– While looking at the maps one can observe that Paris is
more compact and one can walk around much more freely
without the use of transport
– London on the other hand requires the use of transport as
its much larger on the scale.
– In terms of population density Paris definitely outweighs
London by a ratio of 4:1.
– By a recent comparison and taking a look of the most vis-
ited venues Paris definitely has a higher number of restau-
rants of a ratio of almost 3:1 and according to studies
restaurants in Paris have earned higher Michelin Stars than
London’s.
– In terms of Leisure and entertainment London definitely
has more spots than Paris. A simple example would be
that London has more museums than Paris in a ratio of
8:5.
– Paris definitely hosts three of the top 10 most visited at-
traction sites while London has none.
– London definitely has more people from abroad.
– London has a lower temperature than Paris on average.

Artificial Intelligence

Now while comparing which city would be better to start an Ar-

tificial Intelligence company London definitely has more AI compa-
nies.
London has almost 758 AI companies out of which 645 are head-
quatered in the capital.
According to the Mayor of London Sadiq Khan ”“There are few ar-
eas of innovation that have the power to define our future economy
and society more than artificial intelligence”.
According to a recent study london has 868 Number of AI Jobs per
1 Million in the city compared to Paris 660.
One of the major regions this is true is also because of the World-
class Universities which are present in London such as Cambrdi-
ge/Kings/Imperial and so on. London’s status as a global financial

16
Figure 10: Table showing venture capital funding into AI companies across
major global cities from Jan 2013- August 2018

services centre and leader in the development of financial technolo-

gies has helped the city’s AI finance ecosystem to flourish While
in the figure the dataset hasnt been completed , similar trends have
been found in regard to London being one of the top most hub spots
for Artificial Intelligence.
“We would like France to be one of the leaders of AI but we want
Europe to be the champion as well, we have the means and now
we have to create the conditions that will enable us to get there ” -
Emmanuel Macron
Paris however does not fall behind in its capabilities of being an AI
hotspot. Companies such as Facebook, Google, Samsung all intend
to open a Paris AI center and help Paris in its global push into the
AI market.
The goverment has invested almost 1.8Billion Euros for Artificial
Intelligence and promotes that people come and invest in Paris.

17
 Discussion

There are major challenges while constructing a dataset ie:

ˆ The dataset for the Artificial Intelligence wasn’t readily avail-
able and so had to be scrapped from multiple sources which
often leads to inconsistency happening as well as errors.
ˆ Only a random sample of 0.05 percent was taken into consid-
eration. A good and optimal model would take a testing data
and a training data and would train it on the complete dataset
multiple times.
ˆ The data obtained through the API calls would return different
results each time its called. Multiple trials and error runs are
required to get the desired result.
ˆ The districts have too complex geometry which would bring an
error in our analysis if the venues are too close to each other.
This is one of the reason why Pipelines are required.However no
doubt that if this process was to be repeated multiple times the de-
sired outcome would have generated and a better comparison could
have been made.

18
6 - Conclusion
After an indepth review of the comparison between London and
Paris and which city would be a better place to start an Artificial
Intelligence Company or invest multiple conclusions can be drawn.
One of them being that both cities are diverse in their own ways
and boast a culture unlike no other.
Artificial Intelligence is a booming topic and recently more people
have started investing into it as well as companies automating their
processes.
Both cities offer a wide range of opportunities for anyone starting to
invest in Artificial Intelligence or even start a company and various
factors were shown.
Finally a better model could be made by various other methods and
much stronger Machine Learning Algorithms like KD Tree which
have a much faster run time algorithm of O(N log(N )) vs KNN
O(N 2 ).
Furthermore, clustering however did help us to highlight the most
optimal venues and areas.
Finally correlation does not imply causation and so any result here is
subject to change on various other trends and opinions and datasets.

19
7 - Acknowledgements
I sincerely thank all the course instructors who have taken their
time and effort into making this Proffesional Certificate worth the
effort.I also want to state that these are my opinions and are subject
to change as well as I am grateful for all resources and knowledge
that I have learnt throughout this course. I also want to thank God,
my Family as well as friends who have made this a reality and for
supporting me throughout.Thank you to all the peer reviewers that
have graded my projects.

20