Finding a suitable place to open a shopping mall in Delhi.
This is the final project that I have done for the Coursera Capstone Course.
Introduction/ Business Problem
The shopping mall is an emerging and lucrative business in India. However, the business demands the right strategic planning with financial and marketing planning. Shopping malls are getting tremendous popularity these days. And definitely, there are several reasons behind it. According to the retail industry experts, the demand for the mall will more increase in the coming years.
A business article presented that(How to Start a Shopping Mall Business In India — NextWhatBusiness) when you are building a mall or shopping center, entire process starts with market research and ends with marketing and maintenance. First of all, you must do some market surveys or market research and then Finding the location and acquiring the land. Therefore, you should have a proper way to select a land for opening a shopping mall. You have to identify the competition and business opportunities.
Objective
The main goal of this research is to analyze and select the best locations in Delhi to open a new shopping mall. Data science and machine learning techniques like clustering can be used in this case. foursquare location API used to get venue details. Therefore, the project gives the solution to answer the business question: When a businessman looking for a land to a shopping mall in new Delhi, where would you recommend opening it?
Data
For finding an answer to the above problem, we have to use following datasets,
- List of boroughs and neighborhoods dataset with longitudes and latitudes and
- Venue data of shopping malls in the Delhi area (use the foursquare API to find all types of restaurants within a 1000-meter radius for every neighborhood.)
Sources of Data:
The List of boroughs and neighborhoods dataset was taken form Kaggle. (Delhi Neighborhood Data | Kaggle)
This has 4 columns namely, Borough, Neighborhood, latitude and longitude.
The Foursquare Places API provides location-based experiences with diverse information about venues, users, photos, and check-ins. The API supports real time access to places, Snap-to-Place that assigns users to specific locations, and Geo-tag.’(Wikipedia)
Here we are using the explore API call and filtering the search only to find venues that are identified as restaurants.
Methodology
1. Loading Dataset
First, I had to find a neighborhood data set for Delhi. Fortunately, I founded a dataset from Kaggle. I have read that csv file using pandas. There were some NaN values. Therefore, I have dropped rows with null values. These neighborhood data visualized in a map using Folium library.
2. Explore Venues by using Foursquare API
The Foursquare API allows application developers to interact with the Foursquare platform. The API itself is a RESTful set of addresses to which you can send requests. Top 100 venues that are within radius of 2000 meters are retrieved using Foursquare API. There you need to have a Foursquare developer account and obtain the Foursquare ID and Foursquare secret key for making API requests. Then we can make API calls by passing the list of neighborhoods by using a python loop.
Foursquare return venue data in JSON format and we have to extract the venue name, venue category, venue latitude and venue longitude. Then we can check how many venues are returned for each neighborhood and examine how many unique categories can be found.
3. Analyze Venues.
One hot encoding is applied to venues of neighborhood. Each neighborhood is analyzed by taking the mean frequency of occurrences of each venue category. These processing are done to prepare data for K-Means Clustering. Since the problem interested in ‘Shopping Mall’, filtered the neighborhood with ‘Shopping Mall’ venues.
4. K — Means Clustering on Data
we need to cluster all the neighbourhoods into different clusters. The results will allow us to identify which neighbourhoods have a higher concentration of shopping malls while which neighbourhoods have a fewer number of shopping malls. Based on the occurrence of shopping malls in different neighbourhoods, it will help us answer the question as to which neighbourhoods are most suitable to open new shopping malls.
We set the number of clusters to 3 and run the algorithm. After applying the K-Means clustering algorithm, all the neighbourhoods get segregated and form different clusters.Here the Shopping Mall column represents the number of shopping malls in that particular area and Cluster Labels represents the cluster number (either 0 or 1 or 2)
Results
There 125 places in Cluster 1 and that is the cluster with highest number of places. Cluster 0 have 32 places and Cluster 2 having only 6 places. Cluster 0 represents which do not have shopping malls. That means there we can see 32 places which do not have shopping malls, 125 places having one shopping mall and 6 places with 2 or more shopping malls.
The results from the K-means clustering show that we can categorize the neighbourhoods into 3 clusters based on the frequency of occurrence for “Shopping Mall”:
• Cluster 0: Neighbourhoods with very a smaller number of shopping malls
• Cluster 1: Neighbourhoods with a moderate concentration of shopping malls
• Cluster 2: Neighbourhoods with a high concentration of shopping malls
In the following graph Cluster 0 is in red color, Cluster 1 in purple color and Cluster 2 in mint green color.
Discussion
Most of the shopping malls are concentrated to the center of Delhi. You can see there have at least one shopping mall in most of the neighborhoods. Only 32 neighborhoods not having a shopping mall. There is a lack of opportunities to start a shopping mall. But we can select a neighborhood from cluster 0. For that we have to consider more attributes like population of that areas to get a clear decision. But with the available data final conclusion the opportunities to start a shopping mall in Delhi is very less. But this conclusion may vary when considering some more factors.
