/

/

Decision Support for Canine Blood Services: A Platform Architecture for Sparse Veterinary Networks

Decision Support for Canine Blood Services: A Platform Architecture for Sparse Veterinary Networks

Co-authored peer-reviewed IEEE research on decision support systems for canine blood donation networks. Built predictive models on extremely sparse data (176 donors, 482 donations) to forecast demand and donor return patterns. Established architectural foundations for multi-clinic coordination in resource-constrained veterinary settings.

Decision Support for Canine Blood Services: A Platform Architecture for Sparse Veterinary Networks

IEEE Best Paper Award Research: Decision Support for Canine Blood Services
Posted By
Posted At

Uncategory

Posted On

The Problem Nobody Was Solving

Veterinary transfusion infrastructure in India is fragmented. Independent donor panels. Informal WhatsApp groups. No systematic demand forecasting. When a dog needs emergency blood, vets call whoever they know and hope.

I experienced this chaos personally. My own dog needed blood at 2 AM. Four hours of panic. No system. No network. Just desperate phone calls.

That experience became K9Hope. But before building, I wanted to understand if predictive methods could actually help. So I spent a year on research.

What We Investigated

Can you forecast blood demand when data is extremely sparse? Most veterinary networks have fewer than 200 donors. Human blood banks have millions. The methods don't transfer.

We analyzed 482 donations across 176 donors from three referral centers and one regional blood bank. Tiny numbers by human standards. Normal for veterinary settings.

Our approach was conservative. ARIMA-family models for time-series forecasting. Random Forest and logistic regression for donor return classification. Methods that work with limited data and can be explained to clinic staff.

What We Found

Weekly seasonality is real. Monday-Tuesday demand spikes. Weekends drop. Emergency spikes? Unpredictable with available data. The models capture patterns, not chaos.

Donor return classification hit F1 0.77. Modest. Meaningful enough to prioritize outreach, but not enough to automate decisions. We deliberately avoided black-box approaches—veterinarians need to understand why a prediction was made.

The architecture matters more than the algorithm. Multi-clinic visibility. Owner consent tracking. Welfare constraints built into the system. These operational features outperform sophisticated forecasting when data is scarce.

What I Learned

Data sparsity is the default, not the exception. Most real-world problems don't have clean datasets. Building for scarcity requires different instincts than Kaggle competitions.

Interpretability beats accuracy. A model vets don't trust gets ignored. We sacrificed points for transparency.

Ethics constrain design. Canine donors can't consent. Owners can revoke permission anytime. Automated optimization would pressure owners and compromise welfare. We built human-in-the-loop decision boundaries deliberately.

Undergraduates can contribute. This was a student research initiative. We established architectural foundations without clinical deployment. Published in IEEE. Proved that limited resources don't excuse limited ambition.

The Limitations We Owned

We didn't deploy clinically. No veterinary staff evaluated interface designs. Usability claims are speculative. Data interoperability with proprietary practice management systems remains unsolved.

We report these limitations transparently. Research should establish boundaries, not claim false completeness.

Why This Matters

K9Hope became a working platform. This research became a published foundation. Both taught me the same thing: build for real constraints, not idealized assumptions.

The paper is available upon request. The code is on GitHub. The platform is running simulations.

But the real output was learning how to approach problems where data is scarce, stakes are high, and perfect solutions don't exist.

Published in IEEE, February 2026. Co-authored with faculty mentors and research collaborators at Rajalakshmi Institute of Technology, Chennai.

Create a free website with Framer, the website builder loved by startups, designers and agencies.