Page 61 - ISC PROCEEDINGS 21.4

increasingly updated on a periodic basis, contributing to improvements in the ST
transparency index and enhancing oversight by the National Assembly, audit institutions, and
society.
However, the management of PIE continues to reveal several limitations. First,
disbursement progress remains unstable, often slow in the early quarters and
concentrated toward year-end, creating “year-end acceleration” pressure and potential
risks to expenditure quality. The year 2024 is a typical example, when the disbursement
rate reached only about 65% of the plan, reflecting bottlenecks related to land clearance,
investment procedures, and project implementation capacity.
Second, management data remain fragmented. Although systems such as TABMIS,
the public investment information system, and the national e-procurement platform are
in place, these databases have not yet been fully integrated under unified standards;
reconciliation of project, contract, payment, and final settlement data still relies heavily
on manual consolidation. This limits cross-analysis and early risk detection capabilities.
Third, oversight activities are largely ex-post in nature. Inspection and audit
activities are typically conducted according to periodic plans and sampling methods,
making it difficult to comprehensively cover the entire project portfolio. The application
of predictive analytics, ML tools, or anomaly detection techniques in transaction
monitoring remains very limited; data is used primarily for reporting purposes rather than
supporting real-time decision-making.
Fourth, technological infrastructure and human resource capacity have not kept
pace with the requirements of deep digital transformation. Many agencies lack BDA
platforms or specialized analytical centers; financial and treasury officials often lack data
specialists and advanced analytical skills. The legal framework governing data sharing,
data protection, algorithm transparency, and accountability in automated environments
is still under development.
From these limitations, several key causes can be identified: (i) the absence of unified
data standards across systems; (ii) decision-making processes that still rely heavily on
experience and periodic reporting; (iii) limited technological infrastructure and ST allocation
for BDA; (iv) insufficiently synchronized inter-agency coordination mechanisms; and (v) an
incomplete legal framework for data governance and AI application in the public sector.
Overall, the 2021-2025 period reflects Vietnam’s substantial efforts in expanding
the scale of public investment, enhancing digitalization, and improving ST transparency.
Nevertheless, persistent challenges in disbursement, data fragmentation, and ex-post
oversight indicate that the current management model remains more administrative and
procedural than data-driven. This practical context underscores the need to transition
toward an intelligent public investment management model that integrates BDA and AI to
enhance efficiency, mitigate risks, and strengthen accountability in public financial
governance.
3.2. A Big data-based model for monitoring public investment expenditures
The proposed model consists of three organically interconnected components: (i)
data integration; (ii) advanced analytics; and (iii) an executive dashboard. These three
components form a closed-loop cycle encompassing data collection - processing - analysis
- visualization - and decision support.
3.2.1. Data integration
(i) Main Data Sources

60