XccessPoint Proposal

Proposal

Poster

Our Application

Research Paper

The Team

Project Motivation

In recent years, increasing attention has been paid to the issue of inequality in Singapore among parliament discussions and social policy studies. “This is what inequality looks like.” You Yenn Teo’s recent bestseller book (2018) uncovers the heightened tension on social inequalities in Singapore through illuminating an ethnographic presentation of the experiences of the less privileged Singaporeans. Moreover, in the recent Commitment to Reducing Inequality Index 2018 conducted by International Confederation Oxfam International, Singapore was ranked as one of the bottom 10 countries worldwide in inequality reduction. The current state of inequality has motivated to use to particularly delve deeper into the spatial inequality in Singapore which has not been widely researched and examined in the past. We hope to understand inequality by examining the accessibility to many key essential facilities for an ordinary Singaporean living in Housing Development Board (HDB) units. This enables us to highlight the disparity in accessibility between and within neighborhoods in Singapore as geographical accessibility to essential facilities is a key driver for inequality. The improvement in visibility to geospatial inequality through our application could provide policymakers a more justified and structured approach for strategizing future plans in mitigating inequality in different neighborhoods.

Project Description

One way to understand the inequality is to examine the accessibility to many key essential facilities for an ordinary Singaporean living in Housing Development Board units. The aspect of accessibility to look into includes the distance to healthcare facilities (General Practitioner Clinics, Polyclinics and Hospitals), transportation infrastructure (MRT and Bus Stops) , schools, pre-school, police stations, and hawker centres for all HDBs in different planning subzones. We hope to develop an accessibility study tool for urban planners to better strategize the development of new facilities for achieving greater equality for an ordinary Singaporean. For instance, how would Land Transport Master Plan 2040 effectively improve the existing accessibilities to transport facilities. Our team's objective is to analyse and determine how these facilities such as transportation, school and healthcare services would impact the accessibility level around HDB.

Project Objectives

To be filled

Data Sources

Dataset	Description	Data Type	Source(s)
Singapore Regions	To facilitate urban planning, the Urban Redevelopment Authority (URA) divides Singapore into 5 regions, namely Central, West, North, North-East and East Regions.	SHP	Data Source
Singapore Planning Area	Indicative polygon of planning area boundary. To facilitate urban planning, the Urban Redevelopment Authority (URA) divides Singapore into 55 planning areas	SHP	Data Source
Singapore Planning Subzone	Indicative polygon of subzone boundary. The Planning Regions are divided into smaller Planning Areas. Each Planning Area is further divided into smaller subzones which are usually centred around a focal point such as neighbourhood centre or activity node.	SHP	Data Source
HDB	List of HDB location via postal code	CSV	Data Source
School facilities	List of education facilities in Singapore	CSV,KML	Data Source Data Source
Government Markets Hawker Centres	Contains Address of Hawker Centres in Singapore	KML	Data Source
Heathcare Facilities	Contains Address to Healthcare Facilities in Singapore	Website Information	Data Source Data Source
LTA Mrt station	The layer contains the locations of MRT station exits.	KML	Data Source
Bus Stops	All bus stops, bus interchanges, bus terminals in Singapore.	SHP	Data Source

Literature Review

Literature review of relevant research paper on spatial analysis of accessibilities are conducted to enhance our project methodology.

1.Site Suitability Evaluation for Ecotourism Using GIS & AHP: A Case Study of Surat Thani Province, Thailand

Study Objective:

This paper aims to identify and prioritize the potential ecotourism sites using Geographic Information System (GIS) and Analytical Hierarchy Process( AHP) in Surat Thani Province. The factors in consideration for suitability for the land ecosystems include landscape/naturalness, wildlife, topography, accessibility and community characteristics.

Visualization:

Figure 1: Suitability Map for Ecotourism in Surat Thani Province in Thailand

Figure 2:Schematic Representation of the Methodology

Figure 3:AHP Matrix for Pairwise Comparisons and the Consistency Ratio Estimation

Methodology

1.Determination of Weights using AHP

AHP is one extensively used Multi-Criteria Decision Making technique (developed by Saaty in 1980) used in structural decision making process for complex problems that involves multiple criteria across different hierarchical levels. Pairwise comparisons method is used to compare the criteria and allow for evaluation of relative significance of all parameters. Expert opinions were taken into consideration for the comparisons. Pairwise comparison uses a scale of 1 to 9 which 1 means having equal importance while 0 means having extreme importance. Reciprocal pairwise comparisons is used for opposite comparison of facilities.

2.Factoring in Decision Making Inconsistency

To ensure consistent judgement of decision makers, AHP efficiency criteria are measured by Consistency Relationship (CR = Consistency Index/Random Index). If CR is smaller than 0.10, degree of consistency will be fairly acceptable. Otherwise if it exceeds 0.10, inconsistencies exist in the evaluation process and we need to reject the pairwise comparisons and reiterate the process.

3.Land Suitability Assessment

The total suitability score “Si” for each land unit (i.e. each raster cell in the map) was calculated from the linear combination of suitability score obtained for each factor and criteria involved.

where “n” is the number of factors, “Wi” is the multiplication of all associated weights in the hierarchy of “ith” factor ( as seen in Table 5) and “Ri” is a rating given for the defined class of the “ith” factor found on the assessed land unit

Learning Point:

- AHP will be an highly effective methodology for us to reduce the complexity in computing overall accessibility score by structurally factoring the pairwise comparisons of all facilities. Consistency Ratios need to be factor in too.

- Linear weighted combination of accessibility score could be adopted for our study

Caveat:

As this analysis is done on a proprietary software (ArcGIS 9.3), it is difficult for researchers to replicate the methodology of the research paper unless they have access to such software. As we aim to provide urban planners an open-source and easily reproducible application through R programming, there is a need to find similar packages for such methodology on R programming.

2.Estimating Spatial Accessibility to Facilities on the Regional Scale: an Extended Community-based Interaction Potential Model

Study Objective:

The study aims to leverage on measurements of spatial accessibility to regional facilities using aggregated data.The set of facilities includes three types of food outlets on the regional level at Bas-Rhin department,, namely hyper/supermarkets, grocery stores and bakeries.

Visualization:

Figure 4:Maps of number of grocery stores (left) and potential accessibility surface(right)

Methodology

1.Frequency Count of Opportunities within a Given Neighborhood

Frequently used indices based on count was first illustrated to give users a quick overview of the spatial distribution of facilities. This is known as container index which overcomes the limitation of individuals choosing only the nearest facility for consumption and actually consider all available opportunities within a neighborhood.

2. Community-based Interaction Potential Model

This model take into account of difference in urbanization level in the region when computing the accessibility level. As accessibility is a multi-dimensional concept, travel behaviours of the population are factored in. Kernel density estimation and Enhanced Two-Step Floating Catchment Area Method are used for accessibility assessment. These methods consider demand (population) and supply (health practitioners) side as well as travel impedance specification by assigning higher weight to opportunities in nearer region.

Learning Point:

1. Importance of Aggregating Data without consideration of administrative boundaries

To avoid administrative boundaries from limiting the number of closest facilities for each HDB units, administrative boundaries should be ignored when aggregating data. This is more realistic and precise estimation of accessibility levels such that we will not have areas with null accessibility.

2. Data Availability

To effectively implement such customized potential model for spatial accessibility analysis, we need demand, supply of facilities and household characteristics at each HDB units including travel impedance.

Caveat:

1. The study uses population of the region as the base reference to calculate the accessibility to the facilities. However, it failed to take into account of comparison with specific entities (such as HDB flats in our study).

2. The models developed in this study were implemented in the XLISP-STAT programming environment and ArcGIS 9.2 was used for the mapping visualizations. To adopt similar methodology, we need to ensure our programming language have similar functions.

3.Enhancing Healthcare Accessibility Measurement using GIS: A Case Study in Seoul, Korea

Study Objective:

This paper proposes a new method, Seoul Enhanced 2-Step Floating Catchment Area (SESSFCA) to study the accessibility of citizens to healthcare facilities in Seoul. Maintaining accurate and up-to-date information on healthcare accessibility allows the relevant government bodies to strategize future improvements and this includes expansion of healthcare infrastructure, effective labor distribution, alternative healthcare options for the regions with low accessibility, and redesigning the public transportation routes and schedules.

Visualization:

Figure 5:Comparison of Three Methods for Healthcare Accessibility Measurements

Figure 6:Accessibility to Private Healthcare

Figure 7: Box Plot Showing Average Accessibility to Private Healthcare by Districts

Methodology

1.Seoul Enhanced 2-Step Floating Catchment Area (SE2SFCA)

Where S1 and S2 are the standard number of physicians for distinguishing healthcare facilities between a regular hospital, hospital complex and large hospital complex.

This methodology is customized to Seoul city as it factors the fact that the population density is higher and the average distance between healthcare-service locations tends to be shorter than the typical North American or European cities. In addition, Seoul has a higher hospital density than other typical cities. In Korea, a healthcare facility is categorized into regular hospital, hospital complex and large hospital complex in accordance with the size and the number of provided medical specialties. The customized method proposed is more effective and realistic in identifying the regions with weaker accessibility.

2. Critical Distance Boundary Determination

Critical distance boundary (Dt) calculated from the critical travel time, is modeled as a function considering the travel mode of each population. ci is the number of private vehicles per person at population location i, and vc and vp are average speeds of private vehicle and public transportation modes, respectively. In the case of using the public transportation, the travel time boundary is also penalized by subtracting the waiting time (tw) from t. The critical distance boundary sets a distance buffer by factoring in the different socioeconomic status of individuals such as their vehicle possession.

3. Accessibility Calculation

Accessibility to healthcare is determined by geographical distances to service organizations, travel time, available modes of transport, population by region, average car ownership per person, average waiting time for public transportation and Income Differential Indices. The measurement is also separated to accessibility measurement for private and public healthcare facilities.

Learning Point:

To effectively implement such customized potential model for spatial accessibility analysis, we need demand, supply of facilities and household characteristics at each HDB units including vehicle ownership, population breakdown, income level and etc.

Boxplot can be utilized to show attribute distribution of accessibility score

Caveat:

As this analysis is done on a proprietary software(QGIS), it is difficult for researchers to replicate the methodology of the research paper unless they have access to such software. There is a need to source for relevant features/packages in combining location attributes and calculation of important metrics.

Team's Approach

The table below shows a overall summary of our methodology

1. Data Preparation
   • Perform geocoding (Google API)
   • Change coordinate system to SVY21 (EPSG:3414)
   • Merging boundaries on different levels (region, planning area or subzone)
2. Set Accessibility Radius
   • Since comparing across facilities all over Singapore for each house is inefficient and computationally expensive, we will create a buffer of ~1km around each house and calculate distance to facilities within this radius only.
   • If no facilities are found, accessibility radius will be increased to beyond 1km
   • Next, compute minimum distance for each house to each type of facility using Euclidean Distance
3. Accessibility Calculation using Analyic Hierarchy Process
   • User will need to enter his/her priorities for the facility types in a pairwise matrix from -9 to 9
   • Calculate consistency
   Consistency Ratio (CR) = Consistency Index(CI)/ Random Index(RI) CI = (λmax -n) / (n-1) Where λmax is the maximum eigenvalue of the pairwise comparison vector and n is the number of attributes. Random Index is the mean of the resulting consistency index based on the order of the matrix
   • If the criteria entered by user is not consistent(CR>0.1), user will be prompted to reenter the priority
   • If consistent, weight for each facility will be calculated
   • Calculate weighted sum where each weight is multiplied by minimum distance to that facility.
   AHPi = SUM(Wj x Dij) where i is each house and j is each facility. Wj is weight of facility j and Dij is minimum distance from house i to facility j
   • This weighted sum forms the AHP score which is then plotted in the graph. The lower the AHP score, the higher the accessibility

Application Prototype

Data Page

Our team first plans to display the locations of all the facilities in Singapore on a plot. We will also plot a Chrolopleth map where colour of the subzone represents the number of facilities in that subzone

Accessibility Page

Then we will calculate distance from each HDB to the facility selected (i.e. Healthcare Services in this case). We will plot it in a map of the region/planning area/subzone selected. We will also calculate boxplots comparing planning areas and subzones in the region selected.

Analytic Hierarchy Process

Finally, we will calculate AHP for all the houses based on the priority entered and plot it according to the boundaries selected.

Tools and Data Architecture

Technical Challenge

Challenges:

1. Most of the datasets retrieved provided only addresses, not coordinates. Thus, first we had to geocode each point to get the coordinates.

2. Some datasets had CRS WGS84 while some had SVY21. Thus, we had to convert all to SVY21

3. Calculating the distance from each of the 8500 houses to each of the 5000 bus stops was computationally impossible. Thus, we had to use Raster to create a radius around each house and calculate distance from that house to the bus stops which lay within the radius to get the closest bus stop

4. Plotting 8000 points on a map was very cluttered and not insightful. Thus, we provided the user options to select regions/subzones/towns for better plots

Project Task Allocation

Function	Developer	Expected Completion Date	Status
Finalise Storyboard and problem statement	All	Week 9	Done
Collect Data	Raynie	Week 9	Done
Geocoding	Shubham	Week 10	Done
Conversion of CRS systems and merge into boundaries	Shubham	Week 10	Done
Research on AHP calculation	Kaelyn	Week 10	Done
Plot facility distribution	Kaelyn	Week 10	Done
Come up with RShiny template	Raynie	Week 11	Done
Create a buffer of 1km around each house	Shubham	Week 11	Done
Calculate Euclidean Distance	Shubham	Week 11	Done
Calculate Network Distance	Shubham	Week 11	Unsuccessful
Plot boxplots and other comparative plots	Kaelyn	Week 11	Done
Calculate AHP	Kaelyn	Week 11	Done
Add filters in RShiny	Raynie	Week 12	Done
Poster	Raynie	Week 12	Done
Plot facility + house plot	Shubham	Week 12	Done
Beautify maps and RShiny	Kaelyn	Week 12	Done
Update wiki	Shubham	Week 13	Done
Make RShiny user-friendly	Raynie	Week 13	Done
Beautify maps and add additional customisation	Kaelyn	Week 13	Done
Test and debug	All	Week 14	Done
Research Paper	All	Week 14	Done

References

1. https://www.researchgate.net/publication/242450034_A_GIS-BASED_MULTI-CRITERIA_ANALYSIS_APPROACH_TO_ACCESSIBILITY_ANALYSIS_FOR_HOUSING_DEVELOPMENT_IN_SINGAPORE/download
2. https://www.researchgate.net/publication/221354375_GIS-Based_Spatial_Distribution_and_Evolvement_Analysis_of_Urban_Affordable_Housing_A_Case_Study/download

Comments

Feel free to comments, suggestions and feedback to help us improve our project!