DengueOps AI
Simulation-Based Dengue Surge Preparedness Decision Support for Dhaka South
IEEE ICADHI 2026 · Track 06: Health Data Analytics & Predictive Systems
What DengueOps AI Is
DengueOps AI is a simulation-based public health decision-support prototype that converts lag-aware dengue outbreak forecasts into operational preparedness intelligence for Dhaka South, Bangladesh.
It takes a city-level dengue case forecast and produces:
- →Separate empirical forecast-range evidence and preparedness planning scenarios
- →Supply Depletion Horizon (SDH) for NS1/RDT kits and IV fluids per facility
- →Projected bed load and bed gap estimates using LOS approximation
- →Zone priority scores via a spatial exposure heuristic
- →Tiered operational directives per zone and facility
Intended Audience & Role Design
RMSE and model metrics are not intended for public users.
They are included for technical validation and evaluator transparency. Operational users receive translated action recommendations — zone priorities, supply depletion timelines, bed pressure signals, and plain-language directives. The dashboard uses role-based tabs to serve each audience with only the information relevant to their decision-making context.
Operational Command
Public health officials, DSCC vector-control teams, emergency planners
Zone priorities, risk levels, directives, scenario projections
Facility Readiness
Hospital administrators, diagnostic centre managers
NS1/RDT SDH, IV fluid SDH, bed load, bed gap, supply alerts
Public Advisory
Future citizen-facing layer (preview)
Simplified risk status, prevention guidance, when to seek care
Technical Validation
IEEE judges, researchers, MSc evaluators
MAE, RMSE, MAPE, backtest results, feature importance, uncertainty methodology
Operational Workflow
How the system separates data engineering from operational decision-making.
Operational Workflow — Who Does What
Data Sources
External systems
- ·DGHS / IEDCR weekly epi aggregate data
- ·Bangladesh Meteorological Dept. climate data
- ·DSCC facility inventory & bed occupancy feeds
Scheduled Ingestion
MIS / Data Officer
- ·Weekly automated or manual CSV import
- ·Schema validation against data contract
- ·Deduplication and quality flagging
Analytics Pipeline
Technical / Automated
- ·analytics/run_pipeline.py (30–60 s)
- ·Feature engineering → selected Random Forest → RF-bound sensitivity
- ·Operational engine → Directives JSON
Dashboard Alerts
Auto-served to all roles
- ·Pre-computed JSON — zero wait for users
- ·Zone priorities, SDH timelines, bed gaps
- ·Translated action recommendations per role
Human Decision & Action
Hospital / PH / City Corp
- ·Review alerts, not raw data or code
- ·Authorise reorders, bed activation, referrals
- ·Confirm vector-control deployment
Operational users are not expected to code, clean CSV files, or run scripts during an outbreak. The analytics pipeline is maintained by technical/MIS staff, while hospital and public health users receive translated action recommendations.
Future deployment note: In production, the upload/manual CSV mode would be replaced or supplemented by scheduled API/database ingestion from surveillance, climate, facility inventory, and hospital readiness systems. The dashboard and decision layer remain unchanged — only the data ingestion channel is upgraded.
What DengueOps AI Is Not
Framing notice for judges
- ✕NOT a universally suitable forecasting algorithm — Random Forest is the current synthetic demonstration model
- ✕NOT a clinical decision-support tool — no patient-level data, no diagnosis, no triage
- ✕NOT a real-time surveillance system in Phase 0 — static placeholder data only
- ✕NOT a validated operational system — Phase 0 requires further validation before deployment
- ✕NOT an autonomous system — all outputs are advisory and require human review
Why It Is Useful
Fills the preparedness gap
Converts forecast outputs — which are typically unused for supply or bed planning — into actionable operational metrics.
Designed for data-scarce settings
The spatial exposure heuristic and SDH calculation work without ward-level case counts or real-time facility feeds.
Transparent and auditable
All assumptions are disclosed. The system is explainable by design — formula-based, not black-box output.
Human-in-the-loop by architecture
Directives are advisory. The system enhances, not replaces, professional public health judgement.
Portfolio and showcase strength
The end-to-end pipeline — from feature engineering to operational directives — demonstrates applied health data science methodology.
Scalable architecture
The modular Python analytics pipeline is designed to scale: adding new cities, data sources, or model types requires only module replacement.
Future Scalability
Integrate DGHS/IEDCR aggregate surveillance feeds; validate on local ground truth; FastAPI backend
Multi-city deployment; validated spatial model; real-time inventory API integration
Automated directive generation pipeline; health authority integration; Bangla language interface
Calibrated probabilistic forecasting; causal climate-dengue modelling; SDH sensitivity analysis
Core Contribution Statement
“DengueOps AI does not claim a novel forecasting algorithm. Its contribution is the operational decision-support layer that converts lag-aware outbreak forecasts into uncertainty-aware preparedness metrics and public health action priorities.”
Made By
Authors
IEEE ICADHI Project Showcase · Daffodil International University
Meherab Hossain Shafin
AuthorJannatul Tazri Aohona
AuthorAvatar images are placeholders. · DengueOps AI · IEEE ICADHI 2025