REMGIS Proposal

HOME

PROPOSAL

POSTER

APPLICATION

RESEARCH PAPER

Project Motivation

Under the Urban Redevelopment Authority (URA) masterplan that was first released in 2008, URA has been making continuous efforts ever since to grow new employment centres outside the current Central Business District (CBD) in Singapore. This initiative has led to the development of Singapore’s second CBD in the areas of Jurong Lake, together with Jurong East. These joint areas known as the Jurong Lake District, aim to replicate the same amount of success, synergy and vibrancy of the first CBD, or even better. For this to be done successfully, our group aims to conduct thorough analysis and evaluations of the factors that made Singapore’s first CBD the success that it is today.

Project Objectives

This project aims to achieve the following objectives:

1. Analyze the factors that make Singapore’s CBD successful

 a. Identify and study the make-up of professional services within Singapore’s CBD
 b. Examine the landmarks that contribute to the CBD’s success
 c. Understand the factors that influence business locations

2. Replicate and evaluate Singapore’s CBD success factors to the upcoming second business district - Jurong Lake District

Data Preprocessing

Data Source

S/N	Description	Source
1	Legal	http://www.yellowpages.com.sg/search/all/Legal
2	Bank	http://www.yellowpages.com.sg/category/banks
3	Accountancy	http://www.yellowpages.com.sg/search/all/Accountancy
4	Architectural	http://www.yellowpages.com.sg/search/all/Architectural
5	Management Consultancy	http://www.yellowpages.com.sg/search/all/Management+Consultancy

Data Collection

To retrieve the records of all companies in these industries, we performed data scraping on all the companies in the above-mentioned sectors found in the Yellow Pages’ website using R. Once all the companies’ records were retrieved successfully, they were written and saved to a CSV file.

Data Cleaning

One of the problems with the data was that there were some companies that had incorrect postal codes. The incorrect postal codes followed two particular patterns – the first is that the postal code only had 5 digits while the second is that the postal code had a starting postal district which did not correspond to that of Singapore’s. These records had to be manually found by looking through each company’s address and finding postal codes that exhibited either one of the two patterns or both.

After addressing the postal code issue, we realized that for the data to be usable in R, we needed to retrieve the latitude and longitude coordinates of each company. This was done by using the "geocode" function from the "ggmap" package.

Here is a snippet of what the finalized dataset looks like:

Related Works

London’s Central Business District

Goal of study: Examine the range of London’s CBD activities, dynamism and competition of the CBD, the agglomeration of the various types of industries as well as the CBD’s future opportunities.

Kernel density estimation was used to measure the number of professional services’ jobs density per square kilometer. Although we were more interested in the makeup of firms in the CBD and not the number of jobs or employment, we decided that they could build upon this idea and apply kernel density estimation analysis on the various types of firms in Singapore’s CBD instead.

Guangzhou and Shenzhen

Goal of study: Analyze the cartographic definition and representation of the CBD by studying its urban development and functions.

A concentration index is presented to visualize the urban environment by using a density surface that is refined with network distances instead of Euclidean ones. At the end of the study, one of the conclusions of the paper stated that area-based methods of urban analysis such as kernel density estimation are widely used for the purposes of generating a continuous surface with density attribute. This was a positive indication to us to carry on with the initial idea of performing kernel density estimation analysis to analyze the makeup of firms in the CBD and JLD.

Methodology

Location Quotient Analysis

One of the ways to measure success was to determine if the region’s local needs are being met. As a result, we decided to analyse the location quotient values of the five sectors within the CBD and JLD areas, then form our conclusions.

Kernel Density Estimation

To find out about the density of each sector in the CBD, we decided to apply kernel density estimation. This would help to show the density of the various sectors as well as examine which areas of the CBD are more and less dense.

Quadrat Analysis

To determine if firms in the CBD are evenly spaced or clustered, we decided to use a quadrat analysis. We wanted to conduct a test of Complete Spatial Randomness for the 5 industry point patterns, based on quadrat counts before by using the quadrat.test() function of the “spatstat” package.

For the analysis, we will be performing a Monte Carlo simulation with 1000 trails instead of chi-squared tests, as this would obviate the need for all quadrat expected counts to be at least 5 as we noticed in the pre-run that the spatial points of all firms may range from 0 to x number in the CBD window.

Web Application Design

Inspiration

We gathered inspiration for the application’s UI by looking through and using multiple projects that were done by past groups. We took and combined elements that we liked then came up a storyboard for our own application.

Storyboard

Application Architecture

Architecture Overview

Application Layers

Application Overview

Map Tab

Features

• Allows user to upload their own firms data
• The format of the firms data must be in a csv file and in the same format as the data collected from our scrapping script, template of data format can be downloaded for the future.

• Allows user to select the analysis to perform

• Allows user to select the respective sector on the map

• The bandwidth of the Kernel Density plot.
• Smoother plot will be seen with a larger kernel distance while a smaller kernel distance will lead to a plot with more noise.

Spatial Point Analysis

KDE Analysis

Kernel Density Map Comparison Tab

Location Quotient Tab

Features

• Allows user to toggle the respective sector on the map

• Shows LQ figures for both CBD and JLD
• Shows Number of firms in respective regions

• Overview of CBD LQ in Bar Chart

• Overview of JLD LQ in Bar Chart

Quadrat Analysis Tab

Features

• Allows user to toggle the respective sector on the map

• Select respective rows and columns for the quadrat map

All Professional Firms

Selected Firms

Project Timeline

Project Challenges

S/N	Challenges	Solutions
1	Data Scraping API is not readily accessible from Yellow Pages	Explore web scraping techniques in Python and R Lookup Stack Overflow's website for existing samples to test Perform additional data cleaning accordingly for records that are inaccurate, incomplete or inconsistent
2	Geocoding from postal codes to X and Y coordinates to perform thematic mapping	Study external geocoding APIs and adopt one Write a test application to geocode postal codes using the selected API Apply knowledge to project
3	Inexperienced with R Shiny package and R programming	Study R Shiny package and how to use it by going through its documentation and tutorials Self-directed learning on DataCamp​ Refer to past year projects for UI inspiration​ Meet Prof Kam for consultation​
4	Unfamiliar with the implementation of spatial analysis methods such as: Location Quotient K-Means Kernel Density Estimation	Conduct research on R packages which provide the functions that are mentioned Study how similar projects make use of the functions and packages​ Go through documentation and tutorials​ Write test applications to familiarize with the packages​ Consult Prof Kam on major roadblocks​

Project Tools & Technologies

Your comments will be helpful for us to improve our project.

No.	Name	Date	Comments
1.	Insert your Name here	Insert Date here	Insert Comment here
2.	Insert your Name here	Insert Date here	Insert Comment here
3.	Insert your Name here	Insert Date here	Insert Comment here