G1-Group10

Back to Project Home

PROPOSAL

DATA TRANSFORMATION

POSTER

APPLICATION

RESEARCH PAPER

Introduction

International Food Chain (IFC) is a leading brand in its sector, with over 18000 outlets worldwide and an ever-growing presence in the global market. In Taiwan alone, IFC has over 240 branches and are constantly expanding.

However, as the franchise grows bigger, so does its challenges. One of the challenges involves the lack of an analysis to efficiently compare the performance of each chain to one another.

Leveraging on this fact, our group aims to digitalise the data and conduct in-depth analysis on each branch. We hope to track the performance of each chain in relation to Point-Of-Interests surrounding each chain, uncovering and comprehending phenomena, with the aid of spatial data.

Problem and Motivation

To provide an analysis that allows for:

• Digitizing of each chain’s trade and delivery area
• Business profiling of the company’s outlet to determine Points-Of-Interests that can generate insights such as: Highest earning outlets, relative performance of outlets, outlet’s profile patterns and item sales information.
• Allow for informed business decisions, such as determining locations for new outlet openings with matching POIs of high sales outlets
• Scalable program to incorporate future data to generate current information (Using data from other cities besides Taiwan)
• Easy and intuitive tool to quickly view information with regards to all branches

Objectives

This project aims to provide insights into the following:

1. Outlets with the highest monthly sales
2. Relative monthly or yearly performance of each outlet
3. Each branch geographical information, including the type and number of POI’s in the surroundings
4. Profiling of similar types of branches
5. Improvements for poor performing outlets
6. Scalable data processing scripts to incorporate future data

Background Survey of Related Works (WIP)

Visualizations	Explaination
	Data source: https://www.researchgate.net/publication/324949619_Visualization_of_Fast_Food_Restaurant_Location_using_Geographical_Information_System The visualization provides the buffer polygons, as well as representing population density of the area through colour. By comparing the two, we can conclude whether the center of activity is proportional to the population density in a region. It allows us to perform further exploration to see what spatial information significantly affects the level of activity in a city, such as the availability of points-of-interest. This visualization is great as it allows the viewer to clearly see multiple dimensions dealing with spatial data in an elegant way.
	Data source: https://www.researchgate.net/publication/324949619_Visualization_of_Fast_Food_Restaurant_Location_using_Geographical_Information_System The graph on the left shows the distribution of outlets on the geographical map. The right graph describes the outlets grid distribution, result from grid creation and spatial joint operation. From both figures, they can show the potential tendency of whether the outlets are clustered, and with the number of outlets in each grid. We could use them together to justify and adjust the outlet locations.
	Data source: https://www.researchgate.net/publication/324949619_Visualization_of_Fast_Food_Restaurant_Location_using_Geographical_Information_System This visualization provides a novel way of linking a variable to its geographical location when hovering over the respective area. It would be great in our case, if we were to allow the user to view the corresponding branch through the tooltip, for example profit.
	Data source: https://www.researchgate.net/publication/324949619_Visualization_of_Fast_Food_Restaurant_Location_using_Geographical_Information_System This shows kernel density surface, based on the number of fast food restaurants around Jakarta and distribute them smoothly, so it provides average surface estimation. Kernel density estimation allows us to observe both the centrality and agglomeration of existing outlets. This visualization allows us to view multiple dimensions at a time in an effective manner, through the choice of colour and size.
	Data source: https://public.tableau.com/profile/mirandali#!/vizhome/Salesforce-SalesPerformance/SalesPerformance This databoard shows the cumulative sales. We could learn from this and display by outlets to compare the performance by having multiple forms of visualization. We really like the fact that certain key summarizations and variables are displayed on the top, and will consider using this in our project.

Tools and Libraries

The following tools and libraries are used in the digitisation and analysis:

• QGIS
• Python
• R
• Tableau

Datasets

These are the datasets we plan to use:

Dataset	Rationale
Administrative Boundaries, Taiwan	A dataset containing SHP files of the administrative boundaries of taiwan (county, town, village) Used as a reference to digitize IFC branch trade areas
Branch location of IFC, Taiwan	A dataset containing the geographical information of each individual branch. Used as the main target of our project
Point of Interests , Taiwan	A dataset containing each individual Point-Of-Interests in Taiwan (e.g. ATMs, Amusement Parks, Banks) Used as features for analysis with regards to each branch
Powerpoint Slides of trade areas for each branch	A dataset containing each individual hand drawn trade area for each branch in Taiwan Used as features for analysis with regards to each branch
Outlets Monthly Sales Data	A dataset containing the monthly sales information of each individual branch Used to study the sales data along with the profile of each branch to generate yielding patterns (e.g. top and bottom performer)

Foreseen Technical Challenges

We encountered the following technical challenges throughout the course of the project. We have indicated our proposed solutions, and the outcomes of the solutions.

Key Technical Challenges	Proposed Solution	Outcome
Data is already pre-aggregated to display monthly sales	The dataset is given directly to us from IFC, and we are unable to change it. Thus, We shall utilize and do our best with the available data.	NA
Unfamiliarity in R and Python in creation of data processing scripts	Watching video tutorials about R and Python Independent learning on the design and syntax Peer learning and sharing Using Datacamp as our mentor	We managed to start using the languages quickly and suit our own project needs. Each of us work on different parts such as setting up, designing, logic and deployment. This speeds up our project progress.
Data Cleaning & Transformation Proposed Solution	Having a systematic process while working together in order to maximise efficiency e.g. taking turns to clean, transform and perform checks on the data to ensure accuracy	The adopted process was having clear instructions issued to each member in the team, along with maintaining constant communication with each other. In the event that the dataset is deemed too dirty to be usable, it was dropped along with sourcing for new data that would be a suitable replacement.
Lack of geospatial knowledge to understand the dataset initially	Attend SMT201 class to learn more, as well as reading up on resources given by Prof Kam to gain further contextual knowledge	NA
Digitising of trade areas from powerpoint slide to QGIS	The process is manual and we had to put in a lot of effort to convert the drawn polygon to data points in QGIS.	The data points can better allow us to generate insights on the profile of each outlet via its trade area.
Integrating Relevant Data from Multiple Sources Proposed Solution	Working together to decide on what data to extract or eliminate	NA

Roles & Milestones (WIP)

• Roles

Kelvin Chia Sen Wei	Linus Cheng Xin Wei	Eugene Choy Wen Jie
Data Cleaner Wiki Writer	Report Writer Design Architect	Data Cleaner 2 Poster man

• Project Timeline

600px

References

• Project Page: https://wiki.smu.edu.sg/1920t1smt201/GIS_Project
• Python Pandas: https://pandas.pydata.org/
• Tableau: https://www.tableau.com/learn/training
• QGIS: http://www.qgistutorials.com/en/

Comments

Feel free to leave comments / suggestions!