Page 62 - ISC PROCEEDINGS 21.4

First, data from TABMIS include assigned ST estimates, detailed allocations,
expenditure commitments, payments, ST adjustments, and final settlements. This is the
core financial dataset, reflecting actual cash flows and ST obligations incurred by each
project.
Second, the National Public Investment Information System provides information on
total investment capital, funding structure, implementation progress, disbursement
milestones under medium-term and annual plans, and project adjustments. These data
reflect the “physical dimension” of project implementation (physical progress).
Third, the National E-Procurement System stores data on bid package prices,
winning bid prices, savings rates through procurement, and contractors’ participation and
award histories. In recent years, hundreds of thousands of bidding packages have been
published annually, generating a BDA repository for market behavior analysis.
Fourth, audit data and social feedback include audit conclusions of the State Audit
Office, results of sectoral inspections, media reports, and complaint and denunciation
data related to projects. Although largely unstructured, these data sources are highly
valuable for identifying governance risks.
(ii) Building a “digital project profile” for the project lifecycle
Integrating the above data sources enables the creation of a “digital project profile”
for each public investment project. This profile encompasses all information from project
preparation, investment policy approval, capital allocation, procurement, implementation,
payment, and final settlement to audit.
Illustrative example: A transport infrastructure project has a total investment of
2,000 billion VND, implemented over 4 years. Through data integration, the system can
track:
Quarterly disbursement rate compared to the annual plan;
Number of total investment adjustments;
Difference between the bid package price and the winning bid price;
Participation history of the winning contractor in other projects;
Audit recommendations (if any).
If a project undergoes three capital increases within the first two years, records an
abnormal year-end disbursement spike exceeding 40% of the annual plan in December,
and the winning contractor shows a high level of contract concentration within the same
locality, the system can automatically flag it as a potential risk case.
(iii) Data Standardization and Interoperability
To ensure effective integration, it is necessary to standardize project codes, investor
codes, contractor codes, and data classification structures by sector and locality. Such
standardization reduces duplication, improves data quality, and establishes a reliable
foundation for accurate ML algorithms.
Once data is interoperable, management authorities can move from reviewing
fragmented reports to conducting portfolio-level analysis across the entire project
pipeline, thereby strengthening systemic oversight capacity.
3.2.2. Advanced analytics
After data integration, the advanced analytics layer plays a central role in
transforming data into valuable management insights.
(i) Anomaly Detection

61