AMR prediction research work under data scarce condition

As a student at Nazarbayev Intellectual School in Kazakhstan, mentored by a PhD professor in algebra from the Astana IT University(AITU), I am conducting research on predicting AMR risks under data-scarce conditions. Kazakhstan faces challenges with fragmented AMR surveillance, leading to high rates of multidrug-resistant (MDR) and extensively drug-resistant (XDR) infections, especially in bloodstream cases caused by Klebsiella pneumoniae and E. coli.

This study will leverage global proxy data from the Vivli AMR Register to train machine learning models. Specifically, I will:
1. Analyze AMR trends over different years (2015–2023) and by infection type (focusing on bloodstream infections) for ceftriaxone, ciprofloxacin, and meropenem.
2. Preprocess data to classify resistance groupings (MDR/XDR) based on MIC thresholds.
3. Use XGBoost for classification and Prophet for time-series anomaly detection to predict spikes.
4. Generate heatmaps for Kazakhstani regions (e.g., Astana, Almaty) to identify hotspots.

The research will improve patient outcomes by enabling proactive interventions, strengthen antimicrobial stewardship through data-driven guidelines, inform public health by filling surveillance gaps, and bolster health systems in low- and middle-income countries (LMICs) like Kazakhstan. Data will be used solely for non-commercial academic purposes, with results shared in a student journal.