Uncovering Global Flood & Water Security Policy Trends Using Machine Learning, Interactive HTML/Streamlit Tooling & RuleWatcher Primary Source Intelligence
Plan of theme:
Plus
Tagview
Treemap
Transition
CSV Download


AUTHOR
Sumaiya Ibrahim
Graduate student at Kobe Institute of Computing (KIC) Program Analyst II, FCT Emergency Management Agency
1. Key Findings
Algorithmic Policy Classification: Applying k-means machine learning (k=12) to RuleWatcher primary-source data categorized 2,036 official policy records (2019–2026) into distinct macro-themes.
Interactive Tooling: Custom HTML/Streamlit tooling transformed static CSV exports into a live analytics environment for dynamic cluster exploration and parameter tuning.
Dominant Human Dimensions: Social resilience led overall volume, led by Water Scarcity & Drought Management (n=258), Children, Education & Human Rights (n=240), and Displacement & Climate Security (n=206).
Temporal Trends: Multi-year analysis revealed a sustained post-2022 expansion in AI, Governance & Digital Policy, alongside event-driven spikes in public health (2022) and food security (2023).
2. Overview
Tracking global environmental policy requires sifting through thousands of institutional releases, multilateral frameworks, and state regulatory updates. Commercial search engines introduce commercial noise, while academic databases lag active state governance.
This study demonstrates a scalable hybrid workflow: combining RuleWatcher’s verified, structured primary-source database with k-means clustering and interactive web tools. Structuring unstructured policy updates into dynamic thematic clusters accelerates data synthesis and uncovers true baseline trends in global water policy.
3. Description of the Research
This project evaluated global flood and water security policy trends (2019–2026, n=2,036) across four structured phases:
Extraction: Automated retrieval via RuleWatcher using multi-filter Boolean parameters.
Exploratory Mapping: Pre-export metadata inspection using native Tag Networks and TreeMaps.
Machine Learning Execution: URL de-duplication, TF-IDF vectorization, and k-means clustering (k=12) in Python.
Interactive Deployment: Delivery of a Streamlit app and standalone offline HTML/JS dashboard for real-time cluster inspection
4. Global Issue to Be Researched
Flooding and water insecurity are global risks requiring coordinated regulatory responses. While international initiatives call for integrated disaster risk reduction, tracking real-world implementation is hindered by fragmented institutional reporting. Analyzing primary-source policy statements at scale reveals how high-level international commitments translate into concrete action across infrastructure, finance, digital policy, and human rights.
5. How RuleWatcher Was Used
Data was retrieved from RuleWatcher using a unified search filter:
Themes: Climate Mitigation AND Climate Adaptation AND Water Issues
Tag: flood
Timeframe: January 2019 – August 2026 (n=2,036 records)

Figure 1. RuleWatcher Search View displaying primary source document feeds filtered by Theme and Tag parameters.

Figure 2. RuleWatcher Tag View show tags like Plastic Issues, Intelligent Vehicle, and Water Crisis co-occur across global regulations.

Figure 3. Primary interactive TreeMap view categorizing raw keyword frequencies across search results.

Figure 4. Transition Heatmap tracking country-level mention frequency from 2019 to 2026.
6. What Worked Well (and What Didn't) with External Tools
What Worked Well: Ingesting the structured CSV export into a custom Streamlit web interface enabled live k-value adjustments, term-filtering, and rapid AI-assisted metadata tagging.
What Didn’t Work: Standard AI tools lacked intrinsic awareness of RuleWatcher’s native Boolean logic and filter hierarchy. Filter dependencies had to be validated directly within RuleWatcher's interface before importing the dataset into Python scripts.
7. Findings and Usefulness
Topic Cluster Distribution (k=12)
The k-means algorithm grouped all 2,036 documents into 12 macro-clusters, processed via article titles, metadata tags, and text summaries.

Figure 5. Overall article volume across 12 k-means topic clusters (n=2,036).
Multi-Year Policy Trajectories (2019–2026 Trends)
Tracking cluster volume over time revealed key evolutionary shifts:
Emerging Digital Frontier: AI, Governance & Digital Policy shows a continuous upward trajectory post-2022, peaking near 2025 (~40 articles/year) due to increased reliance on smart monitoring and automated risk frameworks.
Event-Driven Spikes: Public Health & Climate Extremes experienced a sharp localized peak around 2022 (~40 articles), while Agricultural Water & Food Security surged in 2023 (~36 articles).
Sustained Baseline: Water Scarcity & Drought Management maintained a high, stable volume (~30–50 articles annually), confirming drought resilience as a perpetual institutional focus.

Figure 6. Small multiples line charts tracking annual article volume per cluster from 2019 through August 2026.
8. Can Other Services Do the Same Thing?
Standard search engines prioritize secondary media coverage, forcing analysts to manually strip away editorial noise. Peer-reviewed academic databases lack real-time institutional statements. RuleWatcher bridges this gap by indexing verified primary policy records with pre-structured metadata fields, providing the clean, standardized input required for custom downstream analytics.
9. Conclusions and Future Perspectives
The overall process proves to be an efficient framework for auditing global policy. Bypassing noise allowed the dataset to reflect true institutional focus areas. Furthermore, building open toolsets demonstrates how raw datasets can be transformed into decision-ready analytics for research teams.
10. Potential for Future Use
My tool is built for other researchers and interns to reuse on their own RuleWatcher CSV exports. Because it runs entirely inside a web browser without uploading data anywhere, the tool offers a simple, private template for practical use—allowing teams to automatically group incoming policy updates by topic or analyze policy trends offline in areas with poor internet access.

Figure 7. Cluster explorer tool, Browser Dashboard Interface (dashboard.html).
11. References
Primary Dataset: RuleWatcher Dataset Export (N=2,036 official policy documents, 2019–2026), exported August 30, 2026.

