Opening A New Restaurant in Canberra Using Machine Learning¶
Business Problem¶
This project will aim to make an educated prediction of what might be the best place to open a new restaurant using the Foursquare API and machine learning learning approach. There are over 200 restaurants in the city of Canberra. So, when opening a new restaurant, one of the most important decision an owner can make is to choose the location wisely.
Canberra is the capital city of Australia and has an estimated population of about 447,692 in the year 2018. According to the 2016 ABS Census data, the median age of residents is about 37 years, and about 1/3rd of the population is under the age of 25 years. So, naturally, a lot of them like to go out during the weekends. Also, Canberra is home to one of the largest universities in Australia (The Australian National University) and other established universities like University of Canberra and Canberra Institute of Technology. These universities attract a lot of international students and academics every year.
So, it's safe to say that Canberra is also culturally rich and ethnically diverse. It's also a crucial decision to decide what cuisine to serve at the restaurant. To help the owner make these decision, I want to use Machine Learning and Foursquare API to come up with a solution that can predict which area/suburb would be the perfect spot to open a new restaurant to maximise footfall. Also, Foursquare API and the Yelp API can be utilised to determine what kind of cuisine is preferred by the people of a suburb/area.
In this way, I think I can make fairly good prediciton for the choice of a restaurant in the city, and this would help maximise the profit of the owner.
Data¶
Solving this problem through machine learning is going to take a lot of data related to the people of Canberra and their food choices. To accumulate this data I'll aim to use one or more of these sources -
Foursquare API¶
The Foursquare API will help me get the data on which suburbs have what type of buildings/dwellings. I will query the API to get the number of each type of buildings in an area and map that out on the map of Canberra. This makes a difference in the overall decision to open the restaurant, as opening a restaurant near office spaces, residential area can be pretty advantageous for business.
Google Places API¶
I can also utilise the Google Places API data to determine the offices and other public places in a particular area/suburb in Canberra. This would provide me with an estimate of which places are more probable locations to have a restaurant clientele. An area with higher footfall but comparitively lesser number of restaurants can be a good place to open a restaurant.
Yelp / Foursquare¶
I also want to utilise Yelp or Foursquare API to get people's reviews on resturants of an area. If a particular cuisine's reviews are not very good in a suburb, then maybe it would be a good idea to open a new restaurant featuring that particular cuisine in that area.
So, in this way, I can make an educated guess as to where to open a restaurant in Canberra.
Methodology¶
Getting Data From Various Sources¶
My first objective was to get the data from various sources and collate them. I found out that Foursquare API didn't have much data related to eating spots for the city of Canberra. So, I went to other sources to fill the gap there. I found that Google Places API would be useful to get this data, especially because it had data for various types of buildings in the city, and, Google is one of the biggest source of GPS and map information.
I collated data from both these APIs and converted the results to a Pandas dataframe for easy readability and analysis. For exploratory analysis, I only needed the location coordinates and names of these restaurants. So, I cleaned up the results to make them look like this -
Although, I didn't have a comprehensive list of all the restaurants and food joints in the city (Google Places API only gives a maximum of 60 results [3 pages of 20 results each] and Foursquare API also gave only 30 results), I had a good enough number to plot these on a map, and visualize what I had got.
Observations¶
From this map, it was easy to observe that -
The suburbs of City and Braddon are densely covered with restaurants and cafes. this is primarily because of the presence of Australian National University in the nearby Acton suburb, and the presence of many government and corporate offices in City area.
Another well-covered region is the Griffith and Kingston suburb of South Canberra due to the presence of many government buildings and lake-side premium real estate. To confirm our hypotheses and get an idea of where all the office buildings and corporate businesses lie in Canberra, it would be a good idea to map out the government and corporate offices (and other places) that may be of interest to a restaurant as potential clientele.
Now, it was time to see which suburbs of the city have more number of offices and other public places that would attract people to restaurants. I utilised the same method to map out the offices and public places in the city and have a look at how they compare to the above map. I thought that it would be a good idea to have both the food joints and the office buildings on the same map (in different colors - green for offices and red for food joints) to visualize the scenario better.
Conclusion¶
So, as we can from the above map, there are two regions of Canberra which have a good number of workplaces,
The suburb of Campbell
The suburb of Parkes
but, both of these suburbs are reasonably untapped in terms of food joints.
- Campbell has a number of schools and sports grounds, so it would be a good educated guess to assume that the area would be regularly visited visited by people aged between 15-30. This age group is drawn towards fast food or casual dining joints more than luxury dining.
- Alternatively, Parkes has a number of museums, government offices and workplaces. So, the visitors here would be more inclined to go to a high-end restaurant to meet colleagues or clients for lunch or dinner. This means, a luxury dining restaurant would be a more rational choice for this suburb.
So, these two locations, in my opinion would be good places to open a new restaurant, depending on what type of restaurant an owner would like to open.