MamaAir/mamaair-kenya-public-dataset

Privacy-preserving synthetic JSON trajectories of 40-week maternal health journeys across climate-stressed Sub-Saharan Africa. Generated from 100+ pregnancy histories in Greater Nairobi, Kenya, it links maternal indicators with climate and air quality data (PM2.5, gases, weather indices) for healthcare AI training and validation. MamaAir Climate-Contextual Maternal Health Dataset. Readme This dataset provides high-fidelity, privacy-preserving synthetic JSON trajectories tracking 40-week maternal health journeys across climate-stressed Sub-Saharan Africa. This data pack was synthesized based on 100+ histories of pregnancy in the region of Big Nairobi (Kenya). It correlates daily physiological indicators with hourly climate and air quality anomalies (Gases, PM2.5, weather indices). Optimized for fine-tuning, validating, and mitigating bioclimatic biases in medical LLMs and healthcare AI applications. ## Official Links - MamaAir AWS public dataset (v1) - MamaAir application on GitHub - MamaAir website Architecture (Data Mining & Processing Pipeline) The end-to-end data mining pipeline transforms decentralized log events into structured longitudinal streams: Ingestion Layer: Captures continuous daily logs from edge sensors (via the B2C mobile application): geo-positioning, tracking behavioural patterns, hydration metrics, symptomatology, and physical workload indicators. Data Enrichment Layer: Automatically maps hourly ambient bioclimatic triggers (Gases, PM2.5 particulates, heat indices, humidity, and UV index) sourced from mix environmental model of OpenAQ, Copernicus Sentinel-5P satellite datasets and ground-level monitoring arrays. Analytical Rules Engine: Dynamically calculates maternal-foetal health risks from environmental data, simulating indoor Household Air Pollution (HAP) exposure curves via domestic fuel, ventilation, and activity schedules. It programmatically generates daily risk-mitigation recommendations spanning four operational domains: diet, behaviour, activity, and mental well-being. Serialization & Tokenization: Aggregates chronological daily logs into weekly gestational slices (weeks 1-40), dynamically mapping metrics against 11 foetal biological development systems. Export (Machine-Readable Format): Generates the final streaming output in JSON (or JSON Lines) format. …