Page 64 - ISC PROCEEDINGS 21.4

Level of delay compared to plan;
Year-end disbursement rate.
For example, logistic regression or gradient boosting models can estimate the
probability that a project will incur irregularities or delays exceeding 20%. Projects with
risk scores above a threshold of 0.7 (70%) would be classified as high-priority monitoring
cases.
If, within a portfolio of 5,000 projects, the system identifies 600 as high-risk,
inspectorates can prioritize these cases, thereby optimizing resource allocation and
reducing monitoring costs.
3.2.3. Smart dashboard
The final layer of the model is the executive dashboard - a data visualization
interface that presents integrated data and analytical results in an intuitive format.
The dashboard incorporates the following functions:
Disbursement rates by sector and locality: Real-time comparison between planned
and actual disbursement.
Plan-actual comparison by quarter/month: Visualization of disbursement trends and
forecasts of annual plan completion.
GIS-based project map: Geographic distribution of projects, with risk levels color-
coded (green - yellow - red).
High-risk project list: Ranked by risk score, capital scale, or degree of delay.
For example, if the dashboard indicates that a province has achieved 85% of its
annual disbursement plan, but 50% of the total disbursement is concentrated in the
fourth quarter, and 30% of its projects fall into the high-risk category, central authorities
can proactively request a review.
Data visualization shortens information lag from several weeks to near real-time,
supporting proactive fiscal management rather than reactive intervention after problems
arise.
Overall Value of the Model
A Big Data-based monitoring model not only enhances risk detection efficiency but
also establishes a foundation for intelligent public financial governance. The three
components - data integration, advanced analytics, and the executive dashboard - form a
closed-loop ecosystem that enables:
Monitoring of the entire project portfolio rather than selective sampling;
Early risk detection instead of ex-post handling;
Optimization of inspection and oversight resources;
Strengthened transparency and accountability.
In the long term, operating this model would contribute to transforming PIE
management from a “procedural control” approach to a “data- and risk-based
governance” model, aligned with the objective of building a modern, transparent, and
digitally adaptive public finance system.
3.3. Comparison of international experiences in applying big data and AI in
monitoring public investment expenditures
The application of BDA and AI in public expenditure oversight has become a key
component of modern public financial governance in many developed countries. Rather
than merely focusing on the computerization of procedures, pioneering nations have
developed interoperable data architectures, integrated ST systems, and risk management

63