Group06 Proposal
Gee-Whiz: Singapore and the suppliers that make her tick
A visual exploration of relationships between government agencies and suppliers.
|
|
|
|
|
|
Contents
Project Overview
Set up in June 2000, the Government Electronic Business (GeBIZ) Portal is an integrated portal for use by all Singapore government agencies to conduct business electronically with their suppliers. As a one-stop center for suppliers to have access to all procurement opportunities in the public sector and to trade electronically with the government, GeBIZ creates an entry point for businesses to access ‘government-to-business’ services.
There are more than S$10 billion worth of business opportunities published annually to 30,000 suppliers registered with GeBIZ. Over the years, GeBIZ has accumulated a knowledge base of procurement data and this massive amount of data can be harnessed as a valuable source of market knowledge.
We will be exploring ways to help visualize the relationships between government agencies and their suppliers, increasing the ability to gain valuable insights. Primarily, we will be exploring network graphs using a R shiny application to help visualize the relationships between the Ministries, Agencies and Suppliers together with other supplementary visualizations.
Goals
Through our exploration of this dataset, below are some of the questions we intend to answer:
- How is government budget currently being spent? Are there certain Ministries with a much higher expenditure than others? Is this consistent or are there large one-off purchases being made?
- Is there any observable trend in terms of government expenditure? (Upward or downward over time for certain ministries?)
- Are there suppliers which are awarded tenders from multiple agencies?
- Which suppliers are more influential in terms of dollar value of contracts awarded? Are these from large one-off contracts from a single agency or from multiple smaller contracts via different agencies?
Aside from our inquiry regarding the dataset, this project also aims to explore R and R packages as a tool for visual analytics. We aim to understand the functionality that is currently available and also to critically evaluate the capacity for future extensibility of the packages used.
Dataset & Preparation
The dataset we will use for our application is sourced from Data.gov.sg
This dataset contains information on successful tenders from January 2015 to December 2017. The structure of the dataset is as seen below:
The process diagram below illustrates our approach to data preparation. The existing dataset only contains agency information. There are no details regarding which ministries these agencies belong to. We will supplement the original data with externally sources information on the mapping. Furthermore, our dataset contains rows for tenders which were not awarded to any supplier. This information is not of interest for our project and will be filtered away. Subsequently, the dataset can be used for preparing the visualizations discussed in the next section.
Visualization Storyboard
There are 4 main types of visualizations we plan to create for our dashboard. These visualizations will help us achieve the goals of answering the questions we have detailed in the section above. For each of the visualization, we start off with the prepared dataset as described above and then further process data into a layout which is appropriate for use.
| Timeseries Chart
Timeseries chart will draw out seasonality patterns in the tenders awarded. The time series charts will be displayed as small multiples in a grid. Each small chart displays the sequential series of tenders awarded by each ministry over the 3 years covered by the dataset. The vertical axis will display the dollar value of each tender.  |
 |
| Treemap
A treemap will breakdown the spend by each ministry into the relevant agencies and then the suppliers. The size of each box is determined by the total dollar value of the tenders awarded while the colour intensity of the chart depicts number of tenders which make up the grid. This treemap will enable us to view the relative spend by ministry and breakdown by agencies.  |
 |
| Boxplot
A boxplot will display the tenders of the top 10 most influential suppliers. Top 10 suppliers are determined by the number of tenders they have successfully been awarded. The dollar value of each of these tenders will be plotted in the vertical axis and a boxplot will display the distribution of these tenders.  |
 |
| Network Graph
A network graph will be used to visualize the size of transactions between government agencies and suppliers. The innermost nodes will be ministries, then agencies and then suppliers. The size of each node will be determined by the dollar value of tenders awarded for each subset of the data.  |
 |
Critique of Existing Work
GeBIZ was initially implemented in 2000 as a single portal for suppliers and government agencies to transact via the tendering process. The development of Business Intelligence (BI) in GeBIZ started in early 2006. GeBIZ BI initiatives can be broadly divided into two areas. The first area entails the development of GeBIZ InSIGHT which is a tool developed internally that leverages Artificial Intelligence (AI) to deliver a set of BI tools, that enable individual procurement users to research historical buys to gain market insights for Intelligent Procurement. The second portion covers the development of GeBIZ Management Console (GMC). GMC enables macro-level portfolio management and performance management in the public sector.
Some key capabilities of GMC include:
- Providing overall spend visibility of government procurement
- Demanding forecasting and identification of potential areas for Demand Aggregation (DA) across the public sector
- Facilitating the establishment of customized procurement Key Performance Indicators (KPI) for individual agencies and the entire government.
However, access to this management console is a paid service which is only granted to Singapore Public Officers for approved agencies. In an effort to democratize access to insights available from this dataset, it is key to develop an open access tool for visualization.
Team GeViz has worked on an application for visualizing the same dataset. This application includes a Treemap, Network Diagram, Sankey Diagram and a Wordcloud. This application makes it difficult to see seasonality trends due to the format in which the data is presented. It is only possible to view one year at a time instead of multiple years which would allow us to understand trends over time. Furthermore, the application also does not allow for understanding of relative weightage of budget allocated to different ministries as there is not visualization to provide an overview of this. All the visualizations only display one ministry at a time.
The Team GeBiz application is another one which has attempted to visualize the same dataset. As part of their project, a Treemap, Sankey Diagram, Calendar Plot and Network diagram has been displayed. A major drawback of the network diagram presented is that it is not sized by transaction value. Thus, though the diagram helps users understand the relationships between ministries, agencies and suppliers, it does not attribute a weightage to this relationship in terms of transaction value. As a result, users are unable to distinguish between the strength of multiple suppliers with a single agency.
The topic modelling techniques discussed by this group are outside the scope of our existing project.
Tools and Packages Used
We primarily intend to use the various R packages detailed below.
Data Manipulation:
- lubridate
- tidyverse
Boxplot:
- ggplot2
Timeseries:
- xts
Treemap:
- treemap
- d3treeR
Network Graph:
- tidygraph
- igraph
- ggraph
- visNetwork
Publishing Shiny Dashboard:
- shiny
- shinydashboard
