The Optimal Location for Indian Restaurant

Introduction

As an Indian and food lover, I was always curious to, what if someone like me want to open Indian restaurant. Great!!!

I choose London for my analysis. London is diverse city. London is home of 500k people roughly 10% of total population, originated from Indian, Pakistan and Bangladesh. India Pakistan and Bangladesh were a single country in colonial era. And it had been ruled by Britain, so lot of people for Indian subcontinent are moved to Britain and especially London, Additional to that, lot of travelers visit to London. Thant’s why I am so excited about London

Business problem

Stakeholder wants to open an Indian restaurant since we have decided to open an Indian restaurant in London, we need to get optimal location. For that we are considering some criteria.

· We will consider the area which has more Indian population

· Restaurant should be not crowed with another Indian restaurants

· It should be as near to city center as possible

Data

The Borough list, Area, Population I get from Wikipedia page [2]. And Indian population I get manually by visiting every borough’s Wikipedia page.

I used geocoder python package [3] to get latitude and Longitude of borough, also to get the distance of borough from city center.

Table 1

Foursquare API is used to get explore the places and get venue and nearby restaurants [4].

I used foursquare category to filter out Indian restaurants [5].

Table 2

Methodology

I used the foursquare API to get top 100 venues at given location. Then I rearrange to get top 10 most common places in borough. Coffee place, Pub, Bar are the most common places in Borough.

Table 3

For further analysis. I used following columns

Table 4

As, this is unlabelled data, we need to use unsupervised algorithms to get cluster of the areas with similar properties like have similar venues and neighbourhood. Then I will compare between the clusters, I used k-means clustering to cluster the data.

Fig -1 shows the different K-values and mean squared distance.

Fig 1

I calculate the Silhouette score for different K, Silhouette score varies from -1 to 1, If it is -1 than clusters are not assigned properly, where data points are assigned to wrong cluster. If its 1 then clusters are perfect.

Fig 2

From fig-1 and fig-2 we can see that k = 4, is well for our data set. So, I cluster the data in four. I used folium library to visualize the data.

Fig 3

Result

Fig-4 show the different borough. Cluster number 3 seems to be more interesting. It has more Indian population and has low number of Indian restaurants Unlike the other cluster, if we narrow down our analysis, we can see that Redbridge has Indian population around 32% of the total population. Redbridge has 27 restaurants in that only one is Indian. Additional to that it is very close to city center around 4.6 Km. That is why Redbridge is very promising area to open the new Indian restaurant.

Table 5

That is why we will choose Redbridge to open a new restaurant. Fig-4 shows all restaurants present in Redbridge borough.

Fig 4

Discussion Limitation and Suggestion

Although we have gone through analysis of boroughs and find out promising borough. But still stakeholder needs to do street level study, like nearby parking area, transportation and connectivity, and area hygiene conditions.

This all thing we cannot find by virtually, stakeholder need to visit places and find location.

Conclusion

London has good number Indian, Pakistani and Bangladeshi people. It is good to open an Indian restaurant in anywhere in London however, the location of the restaurant is key point in business. So, we consider some criteria to open the Indian restaurant.

And according to our analysis, Redbridge borough is very good place to open the new Indian restaurant

References

· [1] London Wikipedia

· [2] List of London boroughs

· [3] Geocoder

· [4] Foursquare

· [5] Developer Foursquare

All the code, details are available in this git hub repo Coursera_Capstone