import pandas as pd

df = pd.read_csv('zomato.csv')

df.head(2)

	Restaurant ID	Restaurant Name	Country Code	City	Address	Locality	Locality Verbose	Longitude	Latitude	Cuisines	...	Currency	Has Table booking	Has Online delivery	Is delivering now	Switch to order menu	Price range	Aggregate rating	Rating color	Rating text	Votes
0	6317637	Le Petit Souffle	162	Makati City	Third Floor, Century City Mall, Kalayaan Avenu...	Century City Mall, Poblacion, Makati City	Century City Mall, Poblacion, Makati City, Mak...	121.027535	14.565443	French, Japanese, Desserts	...	Botswana Pula(P)	Yes	No	No	No	3	4.8	Dark Green	Excellent	314
1	6304287	Izakaya Kikufuji	162	Makati City	Little Tokyo, 2277 Chino Roces Avenue, Legaspi...	Little Tokyo, Legaspi Village, Makati City	Little Tokyo, Legaspi Village, Makati City, Ma...	121.014101	14.553708	Japanese	...	Botswana Pula(P)	Yes	No	No	No	3	4.5	Dark Green	Excellent	591

2 rows × 21 columns

Checking if dataset contains any null

## Checking if dataset contains any null

nan_values = df.isna()
nan_columns = nan_values.any()

columns_with_nan = df.columns[nan_columns].tolist()
print(columns_with_nan)

['Cuisines']

• Cuisines seems to contain null values. Hence any further analysis involving Cuisines the NaN values has to be considered.
• There is an other file which is also available along with this dataset

df1 = pd.read_excel('Country-Code.xlsx')
df1.head()

	Country Code	Country
0	1	India
1	14	Australia
2	30	Brazil
3	37	Canada
4	94	Indonesia

Let us merge both the datasets. This will help us to understand the dataset country wise.

df2 = pd.merge(df,df1,on='Country Code',how='left')
df2.head(2)

	Restaurant ID	Restaurant Name	Country Code	City	Address	Locality	Locality Verbose	Longitude	Latitude	Cuisines	...	Has Table booking	Has Online delivery	Is delivering now	Switch to order menu	Price range	Aggregate rating	Rating color	Rating text	Votes	Country
0	6317637	Le Petit Souffle	162	Makati City	Third Floor, Century City Mall, Kalayaan Avenu...	Century City Mall, Poblacion, Makati City	Century City Mall, Poblacion, Makati City, Mak...	121.027535	14.565443	French, Japanese, Desserts	...	Yes	No	No	No	3	4.8	Dark Green	Excellent	314	Phillipines
1	6304287	Izakaya Kikufuji	162	Makati City	Little Tokyo, 2277 Chino Roces Avenue, Legaspi...	Little Tokyo, Legaspi Village, Makati City	Little Tokyo, Legaspi Village, Makati City, Ma...	121.014101	14.553708	Japanese	...	Yes	No	No	No	3	4.5	Dark Green	Excellent	591	Phillipines

2 rows × 22 columns

Exploratory Analysis and Visualization

Before we ask question on the dataset, it would be helpful to understand the restaurants geographical spread, understanding the rating, Currency, Online Delivery, City coverage…etc.

List of countries the survey is spread across

print('List of counteris the survey is spread accross - ')
for x in pd.unique(df2.Country): print(x)
print()
print('Total number to country', len(pd.unique(df2.Country)))

List of counteris the survey is spread accross - 
Phillipines
Brazil
United States
Australia
Canada
Singapore
UAE
India
Indonesia
New Zealand
United Kingdom
Qatar
South Africa
Sri Lanka
Turkey

Total number to country 15

The survey seems to have spread across15 countries. This shows that Zomato is a multinational company having actives business in all those countries.

from plotly.offline import init_notebook_mode, plot, iplot

labels = list(df2.Country.value_counts().index)
values = list(df2.Country.value_counts().values)

fig = {
    "data":[
        {
            "labels" : labels,
            "values" : values,
            "hoverinfo" : 'label+percent',
            "domain": {"x": [0, .9]},
            "hole" : 0.6,
            "type" : "pie",
            "rotation":120,
        },
    ],
    "layout": {
        "title" : "Zomato's Presence around the World",
        "annotations": [
            {
                "font": {"size":20},
                "showarrow": True,
                "text": "Countries",
                "x":0.2,
                "y":0.9,
            },
        ]
    }
}

iplot(fig)

As Zomato is a startup from India hence it makes sense that it has maximum business spread across restaurants in India

Understanding the Rating aggregate, color and text

df3 = df2.groupby(['Aggregate rating','Rating color', 'Rating text']).size().reset_index().rename(columns={0:'Rating Count'})
df3
df3

	Aggregate rating	Rating color	Rating text	Rating Count
0	0.0	White	Not rated	2148
1	1.8	Red	Poor	1
2	1.9	Red	Poor	2
3	2.0	Red	Poor	7
4	2.1	Red	Poor	15
5	2.2	Red	Poor	27
6	2.3	Red	Poor	47
7	2.4	Red	Poor	87
8	2.5	Orange	Average	110
9	2.6	Orange	Average	191
10	2.7	Orange	Average	250
11	2.8	Orange	Average	315
12	2.9	Orange	Average	381
13	3.0	Orange	Average	468
14	3.1	Orange	Average	519
15	3.2	Orange	Average	522
16	3.3	Orange	Average	483
17	3.4	Orange	Average	498
18	3.5	Yellow	Good	480
19	3.6	Yellow	Good	458
20	3.7	Yellow	Good	427
21	3.8	Yellow	Good	400
22	3.9	Yellow	Good	335
23	4.0	Green	Very Good	266
24	4.1	Green	Very Good	274
25	4.2	Green	Very Good	221
26	4.3	Green	Very Good	174
27	4.4	Green	Very Good	144
28	4.5	Dark Green	Excellent	95
29	4.6	Dark Green	Excellent	78
30	4.7	Dark Green	Excellent	42
31	4.8	Dark Green	Excellent	25
32	4.9	Dark Green	Excellent	61

The above information helps us to understand the realation between Aggregate rating, color and text. We conclude the following color assigned to the ratings:

• Rating 0 — White — Not rated
• Rating 1.8 to 2.4 — Red — Poor
• Rating 2.5 to 3.4 — Orange — Average
• Rating 3.5 to 3.9 — Yellow — Good
• Rating 4.0 to 4.4 — Green — Very Good
• Rating 4.5 to 4.9 — Dark Green — Excellent

Let us try to understand the spread of rating across restaurants

import seaborn as sns
import matplotlib
import matplotlib.pyplot as plt
%matplotlib inline

sns.set_style('darkgrid')
matplotlib.rcParams['font.size'] = 14
matplotlib.rcParams['figure.figsize'] = (9, 5)
matplotlib.rcParams['figure.facecolor'] = '#00000000'

plt.figure(figsize=(12,6))
# plt.xticks(rotation=75)
plt.title('Rating Color')
sns.barplot(x=df3['Rating color'], y=df3['Rating Count']);

Interesting, Maximum restaurants seems to have gone No ratings. Let us check if these restaurants belong to some specific country.

No_rating = df2[df2['Rating color']=='White'].groupby('Country').size().reset_index().rename(columns={0:'Rating Count'})
No_rating

	Country	Rating Count
0	Brazil	5
1	India	2139
2	United Kingdom	1
3	United States	3

India seems to have maximum unrated restaurants. In India the culture of ordering online food is still gaining momentum hence most of the restaurants are still unrated on Zomato as people might be preferring to visiting the restaurant for a meal.

Country and Currency

country_currency = df2[['Country','Currency']].groupby(['Country','Currency']).size().reset_index(name='count').drop('count', axis=1, inplace=False)
country_currency.sort_values('Currency').reset_index(drop=True)

	Country	Currency
0	Phillipines	Botswana Pula(P)
1	Brazil	Brazilian Real(R$)
2	Australia	Dollar($)
3	Canada	Dollar($)
4	Singapore	Dollar($)
5	United States	Dollar($)
6	UAE	Emirati Diram(AED)
7	India	Indian Rupees(Rs.)
8	Indonesia	Indonesian Rupiah(IDR)
9	New Zealand	NewZealand($)
10	United Kingdom	Pounds(��)
11	Qatar	Qatari Rial(QR)
12	South Africa	Rand(R)
13	Sri Lanka	Sri Lankan Rupee(LKR)
14	Turkey	Turkish Lira(TL)

Above table display country and the currency they accept. Interestingly four countries seems to be accepting currency in dollars.

Online delivery distribution

plt.figure(figsize=(12,6))
plt.title('Online Delivery Distribution')
plt.pie(df2['Has Online delivery'].value_counts()/9551*100, labels=df2['Has Online delivery'].value_counts().index, autopct='%1.2f%%', startangle=180);

Only 25% of restaurants accepts online delivery. This data might be biased as we have maximum restaurants listed here are from India. Maybe analysis over city wise would be more helpful.

Let us try to understand the coverage of city

from plotly.offline import init_notebook_mode, plot, iplot
import plotly.graph_objs as go
plt.figure(figsize=(12,6))
# import plotly.plotly as py

labels = list(df2.City.value_counts().head(20).index)
values = list(df2.City.value_counts().head(20).values)

fig = {
    "data":[
        {
            "labels" : labels,
            "values" : values,
            "hoverinfo" : 'label+percent',
            "domain": {"x": [0, .9]},
            "hole" : 0.5,
            "type" : "pie",
            "rotation":120,
        },
    ],
    "layout": {
        "title" : "Zomato's Presence Citywise",
        "annotations": [
            {
                "font": {"size":20},
                "showarrow": True,
                "text": "Cities",
                "x":0.2,
                "y":0.9,
            },
        ]
    }
}
iplot(fig);

<Figure size 864x432 with 0 Axes>

The data seems to be skewed towards New Delhi, Gurgaon and Noida. I see minimal data for other cities. Hence I would do my analysis predominantly on New Delhi.

Asking and Answering Questions

We’ve already gained several insights about the restaurants present in the survey. Let’s ask some specific questions and try to answer them using data frame operations and visualizations.

Q1: From which Locality maximum hotels are listed in Zomato

Delhi = df2[(df2.City == 'New Delhi')]
plt.figure(figsize=(12,6))
sns.barplot(x=Delhi.Locality.value_counts().head(10), y=Delhi.Locality.value_counts().head(10).index)

plt.ylabel(None);
plt.xlabel('Number of Resturants')
plt.title('Resturants Listing on Zomato');

Connaught place seems to have high no of restaurants registered with Zomato, Let us understand the cuisines the top rated restaurants has to offer

Q2: What kind of Cuisine these highly rates restaurants offer

# I achieve this by the following steps

## Fetching the resturants having 'Excellent' and 'Very Good' rating
ConnaughtPlace = Delhi[(Delhi.Locality.isin(['Connaught Place'])) & (Delhi['Rating text'].isin(['Excellent','Very Good']))]

ConnaughtPlace = ConnaughtPlace.Cuisines.value_counts().reset_index()

## Extracing all the cuisens in a single list
cuisien = []
for x in ConnaughtPlace['index']: 
  cuisien.append(x)

# cuisien = '[%s]'%', '.join(map(str, cuisien))
cuisien

['North Indian, Chinese, Italian, Continental',
 'North Indian, Chinese, Italian, American, Middle Eastern',
 'Biryani, North Indian, Hyderabadi',
 'North Indian, European',
 'Cafe',
 'Ice Cream',
 'Continental, Mediterranean, Italian, North Indian',
 'North Indian, Afghani, Mughlai',
 'Chinese',
 'North Indian',
 'Continental, North Indian, Italian, Asian',
 'Asian, North Indian',
 'North Indian, Continental',
 'North Indian, Italian, Asian, American',
 'Bakery, Desserts, Fast Food',
 'North Indian, European, Asian, Mediterranean',
 'Continental, Mexican, Burger, American, Pizza, Tex-Mex',
 'Continental, Italian, Asian, Indian',
 'North Indian, Chinese',
 'Modern Indian',
 'Continental, North Indian, Chinese, Mediterranean',
 'Italian, Mexican, Continental, North Indian, Finger Food',
 'South Indian',
 'Biryani, Hyderabadi',
 'Fast Food, American, Burger',
 'North Indian, Chinese, Continental, Italian',
 'Continental, American, Asian, North Indian',
 'Japanese',
 'Healthy Food, Continental, Italian',
 'Bakery, Fast Food, Desserts',
 'North Indian, Mediterranean, Asian, Fast Food']

from wordcloud import WordCloud, STOPWORDS
import matplotlib.pyplot as plt
import pandas as pd
  
comment_words = ''
stopwords = set(STOPWORDS)
  
# iterate through the csv file
for val in cuisien:
      
    # typecaste each val to string
    val = str(val)
  
    # split the value
    tokens = val.split()
      
    # Converts each token into lowercase
    for i in range(len(tokens)):
        tokens[i] = tokens[i].lower()
      
    comment_words += " ".join(tokens)+" "
  
wordcloud = WordCloud(width = 1500, height = 1500,
                background_color ='white',
                stopwords = stopwords,
                min_font_size = 10).generate(comment_words)
  
# plot the WordCloud image                       
plt.figure(figsize = (8, 8), facecolor = 'orange')
plt.title('Resturants cuisien -  Top Resturants')
plt.imshow(wordcloud)
plt.axis("off")
plt.tight_layout(pad = 0)
  
plt.show()

Top rated restaurants seems to be doing well in the following cuisine

• North Indian
• Chinese
• Italian
• American

Q3: How many of such restaurants accept online delivery

top_locality = Delhi.Locality.value_counts().head(10)
sns.set_theme(style="darkgrid")
plt.figure(figsize=(12,6))
ax = sns.countplot(y= "Locality", hue="Has Online delivery", data=Delhi[Delhi.Locality.isin(top_locality.index)])
plt.title('Resturants Online Delivery');

• Apart from Shahdara locality, restaurants in other locality accepts online delivery.
• Online Delivery seems to be on higher side in Defence colony and Malviya Nagar

Q4: Understanding the Restaurants Rating localities.

• Apart from Malviya nagar, Defence colony in rest of the locality people seems to prefer visiting the restaurants rather ordering food online.
• I would now like to understand the rating of these restaurants that are providing online delivery in Malviya nagar, Defence colony.

• Defence colony seems to have high no of highly rated restaurants but Malviya Nagar seems to done better in terms of Good and Average restaurants.
• As restaurants with ‘Poor’ and ‘Not Rated’ is far lesser that ‘Good’, ‘Very Good’ and ‘Excellent’ restaurants. Hence people in these localities prefer online ordering

Q5: Rating VS Cost of dinning

plt.figure(figsize=(12,6))
sns.scatterplot(x="Average Cost for two", y="Aggregate rating", hue='Price range', data=Delhi)

plt.xlabel("Average Cost for two")
plt.ylabel("Aggregate rating")
plt.title('Rating vs Cost of Two');

I observe there is no linear relation between price and rating. For instance, Restaurants with good rating (like 4–5) have restaurants with all the price range and spread across the entire X axis

Q6: Location of Highly rated restaurants across New Delhi

Delhi['Rating text'].value_counts()

Average      2495
Not rated    1425
Good         1128
Very Good     300
Poor           97
Excellent      28
Name: Rating text, dtype: int64

import plotly.express as px
Highly_rated = Delhi[Delhi['Rating text'].isin(['Excellent'])]

fig = px.scatter_mapbox(Highly_rated, lat="Latitude", lon="Longitude", hover_name="City", hover_data=["Aggregate rating", "Restaurant Name"],
                        color_discrete_sequence=["fuchsia"], zoom=10, height=300)
fig.update_layout(mapbox_style="open-street-map")
fig.update_layout(margin={"r":0,"t":0,"l":0,"b":0})
fig.update_layout(title='Highle rated Resturants Location',
                  autosize=True,
                  hovermode='closest',
                  showlegend=False)
fig.update_layout(
    autosize=False,
    width=1000,
    height=500,)

fig.show()

The aforementioned four cities represent nearly 65% of the total data available in the dataset. Apart from the higly rated local restaurants, it’d be intersting to know where the known-eateries that are commonplace. The verticles across which these can be located are -

• Breakfast
• American Fast Food
• Ice Creams, Shakes & Desserts

Q7: Common Eateries

1: Breakfast and Coffee locations

types = {
    "Breakfast and Coffee" : ["Cafe Coffee Day", "Starbucks", "Barista", "Costa Coffee", "Chaayos", "Dunkin' Donuts"],
    "American": ["Domino's Pizza", "McDonald's", "Burger King", "Subway", "Dunkin' Donuts", "Pizza Hut"],
    "Ice Creams and Shakes": ["Keventers", "Giani", "Giani's", "Starbucks", "Baskin Robbins", "Nirula's Ice Cream"]
}

breakfast = Delhi[Delhi['Restaurant Name'].isin(types['Breakfast and Coffee'])]
american = Delhi[Delhi['Restaurant Name'].isin(types['American'])]
ice_cream = Delhi[Delhi['Restaurant Name'].isin(types['Ice Creams and Shakes'])]

breakfast = breakfast[['Restaurant Name','Aggregate rating']].groupby('Restaurant Name').mean().reset_index().sort_values('Aggregate rating',ascending=False).reset_index()
breakfast.drop("index",axis=1)

	Restaurant Name	Aggregate rating
0	Chaayos	3.812500
1	Starbucks	3.750000
2	Costa Coffee	3.450000
3	Barista	3.325000
4	Dunkin' Donuts	3.300000
5	Cafe Coffee Day	2.573684

import plotly.express as px

df= breakfast
fig = px.bar(df, y='Aggregate rating', x='Restaurant Name', text='Aggregate rating', title="Breakfast and Coffee locations")
fig.update_traces(texttemplate='%{text:.3s}', textposition='outside')
fig.update_layout(
    autosize=False,
    width=800,
    height=500,)
fig.show()

Chaayos outlets are doing better. We need more of those in Delhi. Café coffee day seems to be performing poorly in avg rating. They are required to improve their services.

2: Fast Food Restaurants

american = american[['Restaurant Name','Aggregate rating']].groupby('Restaurant Name').mean().reset_index().sort_values('Aggregate rating',ascending=False)
american

	Restaurant Name	Aggregate rating
0	Burger King	3.477778
3	McDonald's	3.445455
2	Dunkin' Donuts	3.300000
4	Pizza Hut	3.158333
5	Subway	3.047368
1	Domino's Pizza	2.794545

import plotly.express as px

df= american
fig = px.bar(df, y='Aggregate rating', x='Restaurant Name', text='Aggregate rating', title="Fast Food Resturants")
fig.update_traces(texttemplate='%{text:.3s}', textposition='outside')
fig.update_layout(
    autosize=False,
    width=800,
    height=500,)

fig.show()

3: Ice Cream Parlors

ice_cream = ice_cream[['Restaurant Name','Aggregate rating']].groupby('Restaurant Name').mean().reset_index().sort_values('Aggregate rating',ascending=False)
ice_cream

	Restaurant Name	Aggregate rating
5	Starbucks	3.750000
2	Giani's	3.011765
3	Keventers	2.983333
0	Baskin Robbins	2.769231
1	Giani	2.675000
4	Nirula's Ice Cream	2.400000

import plotly.express as px

df= ice_cream
fig = px.bar(df, y='Aggregate rating', x='Restaurant Name', text='Aggregate rating', title="Ice Cream Parlours")
fig.update_traces(texttemplate='%{text:.3s}', textposition='outside')
fig.update_layout(
    autosize=False,
    width=800,
    height=500,)
fig.show()

Foreign brands seems to be doing better than the local brands

Inferences and Conclusions

We’ve drawn many inferences from the survey. Here’s a summary of a few of them:

• The dataset is skewed towards India and doesn't represent the complete data of restaurants worldwide.
• Restaurants rating is categorized in categories
  • Not Rated
  • Average
  • Good
  • Very Good
  • Excellent
• Connaught Palace have maximum restaurants listed on Zomato but in terms of online delivery acceptance Defence colony and Malviya nagar seems to be doing better.
• The top rated restaurants seems to be getting better rating on the following cuisine
  • North Indian
  • Chinese
  • American
  • Italian
• There is no relation between cost and rating. Some of the best rated restaurants are low on cost and vice versa.
• On common Eateries, For Breakfast and Coffee location Indian restaurants seems to be better rated but for Fast food chain and Ice cream parlors American restaurants seems to be doing better.