AY1516 T2 Team HealthTics Analysis final
| MIDTERM PROGRESS | FINAL PROGRESS |
Contents
ABSTRACT
Over the past decade, Health Promotion Board (HPB) has introduced a variety of programmes to increase quality and years of healthy life and prevent illness, disability and premature death in Singapore. Lack of proper tools and a knowledge gap in Geospatial Information System (GIS) analysis has made it difficult for HPB programme planning executives to analyze various facilities at different locations and planning areas.
DIGIVis aims to allow HPB to be more flexible in planning and gathering data-driven insights to facilitate programme planning using a web GIS application. The web GIS application will help to compute KPI of different facilities and visualize results to target users with minimal knowledge of GIS analysis. Programme planners can import additional spatial layers to existing map and perform analysis of various facilities in different communities. Based on these spatial data analysis, HPB can promote better and more effective planning. Moreover,DIGIVis focuses to be intuitive, and user friendly for flexible planning and analyzing.
INTRODUCTION
Health is one of the most concerned topic of the nation where the working stress is increased significantly in the recent years. Singapore government has provided an efficient and widespread healthcare system to ensure that residents enjoy good health with high life expectancy and low mortality rates. With an aim to improve the health status and build a nation of healthy people, Health Promotion Board (HPB) was founded in 2001. Ever since, HPB played the role of the main driver for national health promotion and disease prevention programmes.
In the transition of aligning the board’s vision with the nation’s vision of a smart nation, HPB plans to gather data-driven insights in facilitating its programme planning, HPB utilizes Geospatial Information System (GIS) tools to analyze various facilities in different communities and planning areas. One of the challenges faced in implementing GIS-driven analysis is the knowledge gap and technology cost in the operational level. As such, for a more efficient adoption of GIS applications in the operational level, HPB intends to adopt a lightweight GIS web application to be used by HPB Programme and Outreach Executives which have limited knowledge in GIS analysis.
The system described here, DIGIVis (Dynamic Integrated Geospatial Information Visualization), aims to provide a web-based platform for HPB programme and outreach executives. DIGIVis will handle both spatial and non-spatial data files to retrieve necessary data, compute useful KPI and visualize results to target users. Additionally, DIGIVis allows the users to import additional map layers that can be overlaid together with existing map layers. Such a functionality makes the application more dynamic and scalable to the HPB user needs at the operational level as different requirements of analysis might occur. DIGIVis will also focus on intuitiveness, and ease of use to cater the users with different levels of GIS knowledge.
In this paper, we will discuss on different approaches of implementing GIS services, the architecture and different functionalities of DIGIVis, and comparison with current practice, followed by conclusion.
LITERATURE REVIEW
In the article “A Web-based browsing and spatial analysis system for regional natural resource analysis and mapping” 7, the authors have discussed that traditional GIS techniques and tools are lack of efficiency and ease of use in delivery mechanism. For example, installation or hosting of GIS application and knowledge about geographic and GIS are required to use and understand the GIS analysis. Due to these constraints, many professionals outside GIS community have difficulties in applying GIS techniques in their practical uses. As such, more flexible and feasible web-based GIS applications are introduced.
The authors also explained about earlier versions of server-centric web GIS applications, which consist of 2-tiers: client and server. In these server-centric applications, both data-access and geospatial analysis tasks are carried out in server. The client-tier is the browser and is only used to get access to the server without prior installation in one’s computer. As a result, it overburdens the server over time. Since server cannot handle increasing volume of users and data, such server-centric GIS applications becomes impractical.
Because of performance issues as well as more flexible and advanced programming languages, client-side web GIS are now widely used by professionals in many fields. The flexible programming languages help create a customized environment for the end users so that knowledge about GIS techniques is no longer required in practical applications.
Moreover, the article gave an example of load-balanced web GIS application, WebSAS. WebSAS follows 3-tier architecture: client, application server and database. The architecture spreads out the work load in each tier and the database system provides image files so that the end user do not need understanding about geographic elements. Figure 1. shows the basic architecture of WebSAS’s 3-tiers. The downside of such 3-tiered application is that the cost of setting up each tier is expensive and it requires regular maintenance on application server tier and database tier. Due to high cost of implementation, not many organization could implement web GIS in practical use.
The key factors of web-based GIS application are efficiency and ease of use. Developers should focus on load balancing for efficient data retrieval/access, apply proper GIS techniques, and understand end-user’s requirement from the application.
Review of Similar Work
OneMap is an integrated map system for government agencies to share and deliver location-based services and information among many government agencies. This web-based application is open to public as well as any other organization such as business entities to perform their own geospatial-related analysis. It contains a comprehensive list of map layers such as population density, property market information, etc and their respective metadata.
In addition, OneMap provides aggregated services to display geospatial data on external government websites, such as BizMap, BirdWatching Hotspots, Water Level Sensors. Those websites make use of the base map and some functionalities provided by OneMap as well as secondary sources which are inherited from OneMap. Similar to this functionality, our project could also provide a list of API’s for sharing of base map layer and other basic geospatial functionalities to other HPB websites that require similar services.
By using OneMap, users are able to plot out geospatial data from readily-available services, such as SchoolQuery, LandQuery and PopulationQuery. The concept of this functionality could be applied to our project as well, which is to prepare various datasets that are commonly used by the end-users.
Additionally, users are able to import external KML data to be overlaid on the base map layer together with existing geospatial layer in OneMap. As shown by figure 2 above, layer from RelaxSG.kml denoted by the blue pins and PopulationQuery represented by different colours on the map are overlaid together in OneMap. As such, users are able to analyse the number of parks from RelaxSG.kml together with the population in various planning areas from PopulationQuery. Again, such a functionality could be implemented in our project because it is catered to additional geospatial layers that will be added at an ad-hoc basis by HPB users.
Given the similar functionalities in both our project and those in OneMap, we have observed some limitations in OneMap that should be improved in our project. Some of those limitations are listed below:
- External KML datasets cannot exceed 2MB. Such limitation hinders the overlaying of large datasets.
- Only one KML layer can be imported at a time to overlay on the map as such comparison of different layers cannot be done effectively.
All in all, OneMap is a good starting point for us to have an understanding of a web GIS system. However, it does not have sophisticated analysis functionalities that are required in our projects. As such, we need to do more thorough researches into technical implementations from open-source geospatial libraries, such as TurfJS and GDAL, and modify them to suit our project’s needs.
DATA
DIGIVis uses 3 main data sources to plan out the facility utilization in Singapore. The data sources are 1) base map, 2) HDB map layer and 3) facility map layers.
BASEMAP
OpenStreetMap is used as base map layer in DIGIVis. OpenStreetMap is an open collaborative project to create a free editable map of the world. It is made available as a base map for many other projects. Moreover, this service is one of the base maps used in ArcGIS map viewer.
HDB MAP LAYER
The HDB map layer contains postal codes of all HDBs in Singapore. The postal codes are later converted to SVY21 format to plot on base map layer and perform required calculations.
FACILITY MAP LAYER
Facility map layers contains different map layers of heath promotion options such as park location, smoking cessation touchpoints, healthier dining outlets, etc. These datasets are available to public on data.gov.sg:https://data.gov.sg/dataset?organization=health-promotion-board . The layers can be in non-geospatial files such as csv of HDB with postal codes or geospatial files such as shapefile, and kml files. These data files include information on name of facility, description, location points: coordinate points in spatial file and postal code in non-spatial file, and other related attributes.
PROCESS FLOW CHART
Figure below shows the process flow of DIGIVis. The process starts when users upload layer files of HDB and establishments, including parks and restaurants, to HealthTics system. If the file is non-geospatial, DIGIVis will find a column, such as postcode that can be geocoded, and transform it into a geojson file. If it is a geospatial file, DIGIVis will standardise it into a geojson file.
After created all the geojson files, users can then input filtering specifications, such as “>= 1 fast-food within 200 meter” to find out HDB points that fit the criteria. Based on the distance given by the filter, DIGIVis will create a circular buffer to search for HDB coordinates that are located within the buffer.
HDB points that meet the filtering criteria will be shown into a map. In the final step of the process, users can choose to calculate and display the KPI, such as percentage of residents accessible to the establishments, to the result table. To make DIGIVis more user-friendly, users can choose to hide and show HDB coordinates that fulfil the filtering criteria on the map.
WEB GIS APPLICATION FOR FACILITY PLANNING
The efficiency of a web GIS application largely depends on load balancing framework. Research on difference system architecture design shows that compatibility and complexity are another concern for web mapping applications (Kam, Barshikar & Tan, 2012). Taking into consideration of the latest changes in mobile computing, we design DIGIVis based on JavaScript. To reduce overburdening the server and cost of investment in infrastructure for large servers, our group decides to use client centric web GIS with 2 tiers: Client and Application.
Application tier uses light-weight open source javascript libraries and GIS services to process data and do calculations. Meanwhile, client tier provides functions for user interaction such as file upload, criteria selection and report visualization.
By splitting workload between 2 tiers,DIGIVis avoids overloading only in one tier and works smoothly.
DIGIVis is structured in Express framework which is a light-weight web application framework to easily organize the web application into an MVC architecture on the server side. Along with Express framework, node.js and turf.js are integrated together to perform server-side processing. Node.js is a runtime system allowing to exchange data freely using real-time two-way connections, where both the client and server can initiate communication. To do geoprocessing on web application, DIGIVis uses Turf.js which is a open source javascript library. Turf.js offers useful geoprocessing functions which enables fast and effective data-processing.
Only open-source javascript libraries are used for both front-end and back-end processing since javascript unifies the language and data format (JSON) as well as enables light-weight performance and effective integration with visualization. A number of web services such as ORGE and SLA are also used in DIGIVis to consume spatial functionalities provided by trusted organization and open-source communities. For example, OSM web-service is used to get OpenStreetMap layer.
DIGIVis consists of 3 main steps: 1) file uploading and data preparation, 2) KPI calculation and reporting, and 3) Visualization.
The functions are implemented in 2 web pages as shown in Figure 3 and Figure 4. File uploading and data processing, and KPI calculation and reporting are implemented in page 1 (Figure 3). The final step, visualization, in page 2 (Figure 4) utilizes the results generated from step 2 of page 1.
FILE UPLOADING AND DATA PREPARATION
File uploading functions enables the users to upload their own layer files to perform analysis. As shown in Figure 3, users can upload HDB layer and other health promotion facility layers. The data of HDB layers can be in postal code and that of health promotion facility layers can be in either postal code or XY coordinates. All the data with postal code will be geocoded using web service provided by OneMap Singapore before continuing with projection and conversion to geojson.
In geocoding process, public web service provided by OneMap Singapore is called and get the attributes of each location point. The function is implemented as followed:
In this project, EPSG Projection 3414-SVY21 will be used to perform all spatial functionalities. As such, we can accurately measure distances between location points in kilometres and calculate correct KPI for each buffer in next steps.
Proj4js library used to transform different types of coordinates to 3414-SVY21 coordinate system. Proj4js is an open source javascript library to transform point coordinates from one coordinate system to another.
To use this library, we define Singapore projection as follwed and apply to each of the coordinates of user-uploaded data.
Once the projection is done, the data will be translated into geojson files for its nature of compatibility, lightweight and flexibility. In addition, geojson file has ability to integrate easily with leafletjs and turfjs to generate map-vectors.
To translate the files into geojson format, OGRE web client tool is used. OGRE is a popular ogr2ogr web client(service) to translate various types of spatial data to Geojson format. (The API used: http://ogre.adc4gis.com/convert) This web client is an open source and used in many applications.
For all the uploaded health promotion facility layers, users can choose column to sub-categorize within the dataset.
KPI CALCULATION AND REPORTING
In step 2, users can create queries using query builder in OR and AND conditions. Each selection consists of the health promotion facility/option (or its sub-category), quantity of the facility/option, and buffer distance to calculate KPI. There are two steps in query building: OR query and AND query.
The ‘OR Query Builder’ allows user to choose details of each layer and one or more criteria can be joined with ‘OR’ condition. For each criteria, user has to choose the layer name, operator, number of layer points and radial distance in meters. The selected criteria will be operated in ‘OR’ condition and user has to submit for each combination of ‘OR’ criteria.
For each criteria/set of criteria submitted in ‘OR Query builder’, a new row will be added to ‘AND Query builder’. As such, AND query builder will have a set of ‘OR’ queries and assess them by AND logic. In addition, detailed results will be performed for each OR query and user can toggle to include or exclude the query in final result.
The KPI we will be calculating in this project is the percent of residents with access to health promotion options within x km. The final KPI will be calculated using the following formula:
KPI calculation is based on the query from ‘OR’ and ‘AND’ query builders and pre-processed data. Once a user has selected a query, DIGIVis uses method of turfjs to create a buffer and filter out the HDBs whose coordinates fall inside the buffer. Turfjs creates a buffer from each point of selected layer and filters out the HDBs which fall inside the buffer.
Turfjs is a powerful tool to do geoprocessing on web application. It helps to buffer a location point and then checks whether the desired layers fall within the buffer. Moreover, turfjs offers a lot of useful functionalities and we have used some of them in this project. For example, filter function is used to extract a particular group of data in geojson file. Turfjs API can be found at: http://turfjs.org/static/docs/ For the first step of the query, which is ‘OR’ logic, each HDB id checked if it satisfies any one of the criteria. In second step, ‘AND’ logic checks if all the ‘OR’ logic are satisfied and counts the HDB if it fulfils. After that, total dwelling unit of the filtered HDBs is calculated by summing up the dwelling units of every HDB from the filtered list. Finally KPI is calculated.
KPI is calculated using 2 types of queries: OR and AND query. Firstly, OR query is executed as shown below:
In OR query, buffers are created from each HDB by calling calculateBuffer method which uses turf.js. The calculateBuffer method uses turf.js to create buffer for HDBs as shown below.
After that, calculateBuffer method checks if the points in each layer file fall within the HDB buffers.
Finally, calculateBuffer method checks if the HDBs satisfy the quantity and operator specified by user. A a filtered list is returned to the OR query execution and calculateBuffer method ends.
If it is AND query, the results from OR queries are combined and filtered out the HDBs that fulfil the AND condition.
HDB results are then written into a geojson file so that client tier can access and use in calculating final result and visualizing on dashboard.
The KPI results of each OR query as well as consolidated final KPI of one or more AND queries are then sent back to client tier to include in report. The sub-KPIs, which are the KPIs for OR queries, allows users to understand the breakdown and outliers, if any. Final calculations will be made using AND query and number of households with access and percent of households with access are showed to the user as final KPI reporting. For example, user can measure how many households have access to a healthy dining restaurant within 500 m. This allows HPB, Ministry of Health Singapore and Singapore government to measure and understand current status of each facility provided.
VISUALIZATION
As the last step, interactive visualization features are added to facilitate users understand the data well. For each KPI generated in reporting function, user can save to continue to visualization. Visualization will be implemented using D3.js, and leaflet.js, which are open source javascript libraries. In our visualization as shown in Figure 4, we have used bullet graphs and hexbin maps to transform into effective insights in the interactive dashboard.
The KPIs saved in reporting will be plotted in bullet graph automatically and user can choose the KPI to visualize on the hexbin map. In hexbin map, users can choose different options such as hexbin width, HDB cluster (KPI passed or KPI failed), classification, color, etc. to enhance analysis. User can then click on hexagons to analyze each neighborhood.
BULLET GRAPH
The bullet graphs are used to visualize the overall as well as detailed KPI values. For each KPI, user can set a target value and compare with actual value. For example, when the actual KPI is 76% and target is 80%, users can easily find out the KPIs are not met and sub KPIs will show which area has been affecting the final KPI. In the case when the actual value exceeds target, users can know if they have achieved or set a target in underestimation. We use nvd3.js which provides re-usable charts for d3.js to plot bullet graphs.
Figure 5 shows the bullet graphs for one overall KPI and its sub-KPIs, where users can compare performance of sub-KPIs.
Hexbin Map
Generally, data are displayed in conventional point symbols. However, when the dataset gets larger, point symbol visualization becomes cluttered and inadequate and this can be overcome by using aggregation of data according to specific administrative boundary (Kam, Barshikar and Tan, 2012). Simple way of visualizing aggregated data is to use choropleth map but choropleth map tends to simplify the spatial pattern of the points.
To overcome these deficiencies, DIGIVis uses hexagonal binning on the base map to translate in efficient data aggregation around the bin center. Since hexagon is the polygon with the maximum number of sides for a regular tessellation of a 2D, it is similar to a shape of circle and the hexagonal tessellation becomes perfectly close-packed circles (Nelli, 2014). As such, the hexagonal binning can give most accurate results for our calculations. Many popular spatial analytics tools like ArcGiS also provides hexagon ploygon tool. However, the feature is limited to premium users with ArcInfo license and due to this high cost, the feature is not available to at least 85% of the ArcGIS users (Kam, Barshikar and Tan, 2012).
Due to its usefulness, we implement hexagonal binning to visualize the aggregated data on our map instead of a regular choropleth map. By using aggregated data, the map can show small neighborhood without cluttering point data and tell a better story. On top of hexbin visualization, DIGIVis provides enriched options for the users to adjust the width of hexagon for various level of analysis and choose the data cluster (HDBs fulfilled the query or HDBs dissatisfied the query) to plot on map. Color scheme can also be specified for better visualization for different analysis and individual. The intensity of the color defines the accessibility rate of each hexagon and users can easily differentiate the areas with high and low accessibility rate on the hexagonal binning maps. Finally, the hexagons are clustered in Singapore’s planning zones layer and users can compare different the status of different planning zones.
For detailed visualization, each hexagon can zoomed in and analyzed the neighborhood in detailed map. Detailed map shows the HDB blueprint details of selected hexagon as well as highlights the HDBs that satisfy the user’s selected query for KPI. Therefore, users can find out the areas with low accessibility rate in each small neighborhood.
Leaflet.js is used to visualize these maps and hexagonal binning in this project. Leaflet is the leading open-source javascript library for interactive maps. It is a lightweight javascript library with a variety of mapping features.
COMPARISON WITH EXISTING PRACTICE
At present, Health Promotion Board is using manual practices to do accessibility analysis. QGIS is used to plot different layers on map and get the number of facilities available within a specific buffer size from each HDB. The count of facilities within different buffer size is exported to excel for KPI calculation.
Based the exported data, queries are built and KPI is calculated using excel function. Figure 8 shows the excel query builder HPB is currently using for accessibility analysis. Due to lack of intuitive functionalities in excel, layers are pre-defined and complex functions are required to calculate KPIs.
Therefore, a tedious and manual work is required for perform an analysis. Some technical knowledge is also required to utilize QGIS functions and it is a bit difficult for the HPB’s program planners to perform these analysis due to lack of knowledge in application of QGIS. At the end, it overloads the individuals who can use QGIS and there is only one at HPB for this moment. The only visualization is to use QGIS without in-depth analysis, which requires strong GIS knowledge.
Unlike the current practice of manual analysis, DIGIVis will help the HPB programme and outreach planning executives perform effective analysis without prior GIS background and training. In DIGIVis, each user can use their own layer file without uploading to the application layer or server. This enhances flexibility and freedom of dataset choice to the user. The sub-categorizing function within each dataset allows populating sub layers and precise analysis. In contrast to the current practice, users can build queries dynamically and get detailed KPIs.
With DIGIVis, all the planning executives can visualize the data in interactive dashboard without difficulty. The planning executives can also save and compare different KPIs. Using the interactive bullet graphs, planning executives can analyse each overall KPI and its sub KPIs if they meet the target which can be set individually.
In addition, planning executives visualize different KPI in the map to understand better. The hexagonal binning map allows users to see the distribution of KPI more clearly and precisely. Moreover, users can zoom into each hexagonal area and see details of the neighborhood which falls into a hexagon. Enriched features of map visualization options enables to generate different analysis and visualization.
As such, HPB’s planning executives can perform these analysis and visualization without complicated manual work and GIS knowledge. When compared with existing practice, DIGIVis will save a lot of time and resources, and generate easily understandable and effective data visualization. Using the visualization on the map, planning executives can measure the effectiveness of their previous plans and prepare more effective future plans at different parts/neighborhood of Singapore.
CHALLENGES
In this project, one of the main challenges is to answer Key Performance Indicator (KPI) related queries containing multiple sets of logical operators. For instance, one of the case studies given by the client is related to finding healthier dining options consisting of multiple subcategories of healthier-dining geospatial layers including fast food, restaurant, coffee shop, and others. Each of these facilities is represented as a coordinate point in a geospatial layer. The KPI query may require DIGIVis to find 3 Restaurants OR 1 Fast Food AND 1 Coffee Shop OR 3 Food Courts within 2 kilometre. For our team, this presents a challenging situation where we need to consider the precedence of evaluation for multiple logical operators in the query.
To overcome this issue, we dissect the query input into two parts, which answer the OR and the AND logical operators separately. In addition, we also allow different distance parameters to be entered for each set of statements containing logical operators. Using the example of healthier dining options mentioned above, we require users to enter all the OR statements first. In this case, they need to first enter “3 Restaurants OR 1 Fast Food” within 2 kilometre, then followed with “1 Coffee Shop OR 3 Food Courts” within 2 kilometre. Afterwards, we evaluate the result of these two statements separately then combine the result of both statements with the AND operator to get the final outcome of the query. Users can then repeat the steps to add multiple queries in order to report a KPI.
As mentioned above, there are multiple geospatial layers in a KPI report. Each of this geospatial layer may have subcategory of layers as well. This creates a challenge for our team in terms of making it easy for the users to find the layers that are going to be entered into a query. To resolve this issue, we make use of combobox, which is a combination of text-input and dropdown list. By using a combobox, users can choose an uploaded layer in a list by typing it on the combobox. This allows users to type the layer name efficiently and also to eliminate error of manual typing it in a plain text-input.
In addition, we also challenge ourselves to organize and visualize the KPI report in an effective way for the users to quickly comprehend and analyse the KPI result. In this case, we utilize bulletchart visualization to display each KPI result that is generated by users’ queries. With bulletchart, users can efficiently view the current KPI result as compared to the quota has been set for each KPI. As such, users can view at a glance the KPI that has not reached the predefined quota. To make it even more user-friendly, we allow the users to adjust the quota bar, so as to see how much more they should achieve for a KPI to hit the quota.
When handling geocoding function, we had to choose wisely in terms of asynchronous and synchronous request concepts. In general, a synchronous request will slow down the process while asynchronous request does not store property of data columns. To minimize the tradeoffs and maximize the benefits, we use hybrid approach and attach the property of HDB into geojson without slowing down the process as well as losing crucial data property.
These are a few main challenges we faced in this project. Understanding business process and researching on different approaches of development will definitely help in overcoming these challenges. The open-source libraries and online community also play crucial support in understanding and exploring useful and enriched libraries.
CONCLUSION
The main focus of web GIS applications is to serve flexibility, customizability, and efficiency in spatial analysis. Load-balancing is an important factor to avoid overburdening one layer and pulling down the performance of the application. The cost of basic structure and application can be cut down by avoiding costly servers and using open source libraries/web services. Javascript can be a suitable tool due to its nature of being light-weight, strong community and functionality support from different open-source libraries.
DIGIVis has focused on load-balanced system architecture, flexibility and ease of use for the health promotion and planning executives at HPB. Through DIGIVis, users can easily calculate different KPIs, and analyze the data on effective visualizations. Using the insights and information from this analysis, DIGIVis can facilitate health promotion and programme planners in their future plans.
REFERENCES
StraitsTimes. (2014, May 7). One-quarter of workers report high stress levels. Retrieved from www.straitstimes.com/singapore/health/one-quarter-of-workers-report-high-stress-levels OneMap. (n.d.). OneMap. Retrieved from http://onemap.sg
Singapore Land Authority. (2015, October 29). About OneMap. Retrieved from http://www.sla.gov.sg/e-Services/OneMap.aspx#.VpJjzcZ97IU
EnterpriseOne. (n.d.). BizMap. Retrieved from http://apps.enterpriseone.gov.sg/Bizmap/BizmapEnterpriseone.aspx
Nature Society of Singapore. (n.d.). Bird Watching Locations. Retrieved from http://www.nss.org.sg/wildbirdsingapore/Default.aspx
Public Utilities Board of Singapore. (n.d.). Managing Flash Floods. Retrieved from http://www.pub.gov.sg/managingflashfloods/WLS/Pages/WaterLevelSensors.aspx
Vatsavai, R. R., Burk, T. E., Wilson, B. T., & Shekhar, S. (2000). A Web-based browsing and spatial analysis system for regional natural resource analysis and mapping. Proceedings of the Eighth ACM International Symposium on Advances in Geographic Information Systems - GIS '00.
Smith, N. (2012, May 25). Binning: An Alternative to Point Maps. Retrieved March 27, 2016, from https://www.mapbox.com/blog/binning-alternative-point-maps/
P. Singh, D. Chutia and S. Sudhakar, "Development of a Web Based GIS Application for Spatial Natural Resources Information System Using Effective Open Source Software and Standards," Journal of Geographic Information System, Vol. 4 No. 3, 2012, pp. 261-266. doi: 10.4236/jgis.2012.43031.
KAM, Tin Seong; BARSHIKAR, Ketan; and TAN, Shaun Jun Hua. DIVAD: A Dynamic and Interactive Visual Analytical Dashboard for Exploring and Analyzing Transport Data. (2012).International Journal of Computer, Electrical, Automation, Control and Information Engineering. , 6(11) , 834. Research Collection School Of Information Systems. Retrieved from http://ink.library.smu.edu.sg/sis_research/1761
Nelli, F. (2014, January 14). Hexagonal Binning – a new method of visualization for data analysis. Retrieved from http://www.meccanismocomplesso.org/en/hexagonal-binning/
