INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)

ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025

Page 3768

Smart Budget Allocation in Public Policy: A Data-Driven Approach

for Equitable Resource Distribution

Ashok Teja Kaki., Dr. K Srikanth., T.Venkatesh., Tharun kumar

IT Department, JNTU GV College of Engineering, Vizianagaram, India

DOI: https://doi.org/10.51244/IJRSI.2025.120800336

Received: 23 Sep 2025; Accepted: 29 Sep 2025; Published: 13 October 2025

ABSTRACT

Public policy implementation often struggles with uneven budget allocation across sectors and regions, leading

to inefficiencies in resource use. This study presents a data-driven framework for smart budget allocation

through predictive analytics and optimization methods. The proposed model helps policymakers ensure fair

and efficient distribution of public funds by integrating socioeconomic indicators, sector-specific requirements,

and past expenditure outcomes. Using linear regression forecasting combined with constrained linear

programming, the framework determines sector-wise budgets. The analysis focuses on five crucial public

sectors—Agriculture, Health, Education, Rural Development, and Road Transport—using data from 2020 to

2025, along with macroeconomic indicators such as GDP growth, unemployment rates, and simulated public

sentiment. A custom interactive dashboard enables real-time visualization and engagement with predicted and

optimized budgets. Evaluation results highlight the potential of blending machine learning with operations

research for evidence-based governance. The study introduces a scalable and reproducible model that aligns

with national missions like India@2047 and Swarna Andhra Pradesh. By embedding data science into fiscal

decision-making, this work contributes to advancing digital governance, improving transparency, and fostering

citizen-centric planning.

Keywords: Budget Allocation, Public Policy, Resource Optimisation, Data-Driven Governance, Predictive

Analytics, Policy Modelling

INTRODUCTION

Turning development goals into reality requires more than vision—it calls for smart, evidence-based budget

allocation. The Union Budget is the government’s key tool for directing resources to priority areas. Studying

allocation patterns from 2009 to 2024 helps reveal shifts in policy focus, sectoral priorities, and alignment with

broader national and state-level goals.

This research looks at how budget allocations reflect trade-offs, priorities, and their consistency with long-term

visions such as India@2047 and Swarna Andhra Pradesh, both of which aim to make India a developed,

inclusive, and digitally empowered nation by mid-century. While much has been written on in public finance

in India, most studies focus only on overall spending trends. Few explore sector-level efficiency, equity, or

strategy. This study fills that gap using a data-driven approach to identify underfunded sectors and evaluate

alignment with policy goals.

In public finance in India, most studies focus only on overall spending trends. Few explore sector-level

efficiency, equity, or strategy. This study fills that gap using a data-driven approach to identify underfunded

sectors and evaluate alignment with policy goals.

Globally, countries like the UK and Singapore use advanced forecasting and real-time analytics for

governance, but India still relies largely on static reports. This leads to inefficiencies and biases in allocating

funds, particularly in underserved areas like rural development and education. Bridging this gap requires AI

systems that are transparent, inclusive, and supportive of human decision-making.

INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)

ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025

Page 3769

This study introduces a unified framework that combines prediction, optimization, and impact visualization. Its

contributions include:

1. A hybrid budgeting model integrating machine learning and optimization.

2. Sectoral impact analysis using sentiment and macroeconomic data.

3. An interactive decision-support dashboard for participatory governance.

4. A policy-driven evaluation model aligned with India@2047 and Swarna Andhra Pradesh.

Problem Statement

India’s long-term visions, such as India@2047 and Swarna Andhra Pradesh 2047, aim to create a developed,

inclusive, and future-ready economy. But achieving these goals depends heavily on how budgets are

allocated—both fairly and efficiently. While annual budgets form the backbone of public spending, they often

fail to align with long-term priorities, especially in critical sectors like infrastructure, healthcare, education,

and agriculture. Although past studies have examined India’s budget patterns, most remain broad and overlook

sector-specific trends over time. This leaves a key gap in understanding how allocations from 2020 to 2024

reflect the government’s stated priorities at both national and state levels.

To bridge this gap, this study focuses on:

Analyzing sector-wise central government budget allocations from 2020 to 2025.

Assessing how well these trends align with development goals under India@2047 and Swarna Andhra Pradesh

2047.

Identifying underfunded sectors or mismatches between vision and fiscal planning.

Recommending strategic improvements for more inclusive and results-oriented budgeting

METHODOLOGY

The proposed smart budget allocation framework integrates budget prediction and optimisation using publicly

available data, economic indicators, and public sentiment analysis. The implementation involved six key

stages:

Data Collection and Preparation

Budget data for five key sectors—Agriculture, Health, Education, Rural Development, and Road Transport—

was collected from India’s Union Budget documents (2020–2025). Public sentiment scores and

macroeconomic indicators such as GDP growth and unemployment were also included.

Dataset Enrichment

Each sector’s data was enriched yearly with GDP growth (%), unemployment (%), and a simulated public

sentiment score (0–20 scale) reflecting demand or importance

Predictive Modelling

Linear regression models were trained separately for each sector using year, GDP growth, unemployment, and

sentiment scores. These predicted sectoral budget allocations for 2026. Accuracy was tested using Mean

Squared Error (MSE) and R² scores.

INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)

ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025

Page 3770

Optimisation

A Linear Programming (LP) model (via PuLP) was used to allocate budgets effectively under overall

constraints. Objectives included maximizing public satisfaction, economic growth, and unemployment

reduction. Logarithmic utility functions simulated diminishing returns, with sectoral limits set to reflect policy

rules.

Visualisation and Dashboard

Results were shown using bar and line charts. A custom Streamlit dashboard enabled interactive exploration of

allocations, trends, and predicted vs. optimized budgets.

Evaluation Regression

models were evaluated for prediction accuracy, while the optimisation model was assessed by its ability to

meet priorities within budget limits. Sectoral alignment and public priorities were qualitatively validated.

Fig 1. Flow Chart

Results and Interpretation

This section presents the outcomes of both the predictive (regression) model and the optimisation model

developed to allocate government budgets across five major public sectors: Agriculture, Health, Education,

Rural Development, and Road Transport

Fig 2. Total (₹ Cr) vs GDP Growth (%)

INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)

ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025

Page 3771

Fig 3. Revenue

Fig 4. Sector vs Total (₹ Cr)

Predicted Budget for 2026

Using regression models trained on 2020–2025 data, budget allocations for 2026 were forecasted across five

sectors. The models incorporated GDP growth, unemployment, and public sentiment, achieving strong

accuracy with R² scores between 0.88 and 0.94. This indicates a reliable relationship between these factors and

budget outcomes.

Optimisation Results

An optimisation model was applied to distribute ₹7,00,000 Cr across the same sectors. It aimed to maximize

overall utility by balancing public sentiment, economic growth, unemployment reduction, and diminishing

returns on excess funds. Sectoral limits ensured compliance with policy priorities. Results showed close

alignment with predicted needs, while shifting funds strategically—such as boosting Agriculture allocations to

improve public satisfaction.

Visual Insights

Bar plots compared predicted vs. optimised budgets, showing broad alignment with minor adjustments across

sectors. A real-time Streamlit dashboard further enabled interactive exploration of year-wise and sector-wise

trends.

Ethical and Policy Implications Using AI in public budgeting raises concerns of fairness, transparency, and

bias. While the models here use interpretable methods, they rely on simulated sentiment scores and general

economic indicators. Therefore, the framework should act as a decision-support tool, not a decision-maker.

Responsible use requires human oversight, clarity of inputs, and accountability mechanisms.

INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)

ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025

Page 3772

Table 1: Summary of Budget Predictions, Optimisations, and Model Metrics

Sector

Sentiment Score

Predicted (₹

Cr)

Optimised

(₹ Cr)

% Change

R² Score

Priority

Rank

Agriculture

1,13,153

1,30,000

+14.9%

0.91

Health

89,000

0.0%

0.88

Education

1,24,000

1,22,000

-1.6%

0.94

Rural Development

1,65,000

0.0%

0.93

Road Transport

2,75,000

2,65,000

-3.6%

0.92

Figure 5: Dashboard

Comparative Review with Existing Studies

Table 2 : Comparative Reviews

Study

Method Used

Scope

Limitation

This Study’s Contribution

Makwana

(2024)

Text Analysis

Union Budget

Lacks predictive

modeling

Introduces regression and linear

programming (LP) models

Radulescu

(2015)

Neural

Networks

EU Public

Spending

Ignores policy

alignment

Integrates forecasting with national

policy roadmaps

This Study

Regression + LP

Sectoral India

Simulated

sentiment only

Aligns with India@2047 vision and

offers open-access dashboard

CONCLUSION

This study introduced a data-driven framework for smarter budget allocation by combining predictive analytics

with linear programming. Using historical budget data, macroeconomic indicators, and public sentiment, the

model successfully forecasted sectoral needs and generated optimized allocations within fiscal limits. The

results demonstrated that integrating sentiment and economic context can make resource distribution more

equitable and strategically aligned, while optimisation modelling highlighted budgetary trade-offs and

improved fairness in allocation.

INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)

ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025

Page 3773

Future enhancements could include testing advanced machine learning approaches, such as random forests or

neural networks, to improve long-term forecasting. Scenario-based simulations—covering fiscal shocks,

pandemics, or climate change—could strengthen resilience, while stronger ethical safeguards, including

explainable AI and accountability mechanisms, would improve transparency and trust. A comparative analysis

with global best practices, such as those in the UK and Singapore, would further contextualize the Indian case

and position the framework within international discussions on AI-enabled public finance.

Future Scope

This work provides a foundation for data-driven budgeting but leaves room for growth. Future improvements

could include expanding the model to state or district levels, using real-time economic data, and refining public

sentiment analysis through social media and surveys. The system can also be enhanced for multi-year planning

and advanced visualisation with tools like Power BI or Tableau. These extensions would further strengthen

alignment between fiscal planning, ground realities, and long-term policy goals.

DISCUSSIONS

Table 3: Future Advanced Evaluation Metrics

Metric

Description

Purpose

Status

Sensitivity Analysis

Impact of ±1% change in GDP, sentiment, and

unemployment

Model robustness

Planned

Historical Validation

Compare model output for 2024 with actual allocations

Accuracy check

Planned

Stakeholder Impact

Estimation

Link budget shifts to outcomes (jobs, health)

Policy relevance

Planned

Ethical Risk Mapping

Fairness and bias audit for decision support

Responsible AI

governance

Added

Scenario Testing

Simulate high vs low economic growth impacts

Strategic foresight

Planned

REFERENCES

1. Ministry of Finance, Government of India. (2024). Union Budget Documents (2020–2025).

https://www.indiabudget.gov.in

2. NITI Aayog. (2022). Strategy for New India @ 75. https://www.niti.gov.in

3. Pedregosa, F., et al. (2011). Scikit-learn: Machine Learning in Python. JMLR.

4. PuLP: A Linear Programming Toolkit for Python. https://coin-or.github.io/pulp

5. Streamlit: Build Data Apps in Python. https://streamlit.io

6. Public sentiment and economic indicators were simulated for academic use in the absence of official

datasets.

7. Makwana K (2024) A Textual Data Analysis of the Union Budget of India. Indian Journal of Science

and Technology 17(5): 478-486.

8. Radulescu Magdalena, Banica Logica & Tatiana Zamfiroiu (2015) Forecasting public expenditure by

using feed-forward neural networks, Economic Research Ekonomska Istraživanja, 28:1, 668-686, DOI:

10.1080/1331677X.2015.1081828

9. Government of Andhra Pradesh. (2025). Swarna Andhra@2047 Vision Plan. Retrieved from

https://spsnellore.ap.gov.in/document/swarna-andhra2047-vision-plan/

10. Radulescu, M., Banica, L., & Zamfiroiu, T. (2015). Forecasting public expenditure using feed-forward

neural networks. Economic Research, 28(1), 668–686.

11. Pedregosa, F., et al. (2011). Scikit-learn: Machine Learning in Python. JMLR, 12, 2825–2830.

INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)

ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025

Page 3774

12. Winston, W. L. (2004). Operations Research: Applications and Algorithms. Thomson Brooks/Cole.

13. PuLP: A Linear Programming Toolkit for Python. https://coin-or.github.io/pulp

14. UN ESCAP. (2020). AI for Social Good: Asia-Pacific Perspectives. https://www.unescap.org

15. OECD (2019). Embracing Innovation in Government: Global Trends 2019. https://oecd-opsi.org

16. Floridi, L., & Cowls, J. (2019). A Unified Framework of Five Ethical Principles for AI in Society.

Harvard Data Science Review.

17. Jobin, A., Ienca, M., & Vayena, E. (2019). The Global Landscape of AI Ethics Guidelines. Nature

Machine Intelligence, 1(9), 389–399.

18. Sunstein, C. R. (2003). The Cost-Benefit State: The Future of Regulatory Protection. ABA Journal

19. Srivastava, A., & Rai, A. (2022). Using Data Analytics for Public Sentiment in Policy Making. Journal

of Policy Modelling, 44(3).