This research arrives at a pivotal moment for Indonesia, where government agencies manage vast amounts of "Big Data" but often struggle to translate that information into effective, targeted interventions. By using an "applied ensemble" approach, the authors provide a roadmap for more efficient social assistance and healthcare programs.
Navigating Complexity in Public Policy
Social issues like stunting are rarely the result of a single factor. They are deeply intertwined with family income, parental education levels, and regional access to healthcare. Traditional statistical methods often overlook the non-linear and hidden patterns within this data, leading to policies that may not reach those most in need.
The researchers at Universitas Amikom Yogyakarta recognized that to solve these "wicked problems," policymakers need tools that can handle "heterogeneous data"—information that varies wildly from one region to another. The goal of the Antariksa and Karimah study was to create a system that is not just technically advanced but also practical and interpretable for government officials who may not be data scientists.
The Methodology: A "Team" of Algorithms
The core of the research is the "Applied Ensemble Machine Learning Framework." Rather than relying on a single computer model, an ensemble framework combines the strengths of multiple algorithms to produce more stable and accurate predictions.
Muhammad Maulana Antariksa and Shofiyati Nur Karimah utilized two primary supervised learning models:
- Random Forest: This model operates like a large group of decision-makers, where each "tree" provides a prediction based on different data subsets, and the majority vote determines the final result.
- Gradient Boosting: This model learns from previous errors, step-by-step, to refine its accuracy over time.
By merging these models, the framework can approximate the complex relationships between input variables—such as household income and maternal education—and outcome indicators like child nutritional status.
Key Findings: Precision and Stability
The experimental results from the Universitas Amikom Yogyakarta study highlight the superiority of ensemble learning over traditional models. Key findings include:
- Higher Predictive Accuracy: The ensemble framework achieved significantly higher accuracy in predicting welfare indicators compared to using a single algorithm.
- Feature Importance Identification: The system successfully ranked variables, finding that family income and the educational background of mothers were the strongest predictors of community health outcomes.
- Robustness Across Regions: The model remained stable even when tested against diverse datasets, suggesting it can be used effectively across different provinces in Indonesia.
- Interpretability for Officials: Unlike "black box" AI systems, this framework emphasizes feature importance analysis, allowing policymakers to see exactly which factors are driving the predictions.
Societal Impact and Better Governance
The implications of this research for governance and social welfare are profound. With this framework, government agencies can move away from a "one-size-fits-all" approach to social assistance. For example, in the fight against stunting, the AI can help identify specific districts where educational interventions for mothers would be more effective than simply providing food aid.
Beyond healthcare, the Universitas Amikom Yogyakarta researchers suggest that this data-driven approach can be applied to urban planning, poverty reduction, and emergency response. By identifying high-risk populations before a crisis hits, the government can allocate resources more proactively.
As Muhammad Maulana Antariksa notes in the study, the framework is built for real-world application. "The proposed framework emphasizes robustness, interpretability, and practical applicability rather than algorithmic novelty," ensuring that technology serves as a reliable bridge between raw data and meaningful social change.
This study positions Universitas Amikom Yogyakarta at the forefront of "AI for Social Good," proving that the smart use of technology is essential for building a more equitable and healthy society.
Author Profiles
- Muhammad Maulana Antariksa, B.Comp.Sc. (Cand.): A researcher at the Department of Computer Science, Universitas Amikom Yogyakarta. His expertise lies in decision support systems, predictive analytics, and the application of machine learning to social science.
- Shofiyati Nur Karimah, M.Kom.: A faculty member and researcher at Universitas Amikom Yogyakarta. Her research interests include advanced data analysis, ensemble learning, and computational intelligence for public policy.
Source Research
Article Title: Applied Ensemble Machine Learning Framework for Data-Driven Decision Support Using Socioeconomic Data
Journal Name: Formosa Journal of Multidisciplinary Research (FJMR)
Publication Year: 2026
Volume/Issue: Vol. 5 No. 1, pp. 275–284
Official URL:
0 Komentar