OUR APPROACH: STETA
Our Space-Time Encoded Training Architecture (STETA) embeds spatial and temporal relationships into training data as scalar values, allowing standard machine learning models to recognize patterns that would otherwise remain hidden.
Most spatial models address complexity by modifying the algorithm itself. STETA embeds the intelligence upstream, in the training set — so instead of learning from isolated data points, the model learns from meaningful groups of related information. The result is outputs that are more robust, more realistic, and more useful in practice.
WHERE WE APPLY IT
GeoNexa supports decisions wherever location, time and local context affect risk, demand or performance.
WHY GEONEXA LABS?
Real-world risk does not follow neat boundaries
GeoNexa models the relationships between places, networks and time so local signals are not lost in broad averages.
Our work is underpinned by 30 years of spatial data science research, published in Geographical Analysis, Annals of the AAG, Computers, Environment and Urban Systems and other leading international journals, and presented at GEO Business 2025 and AGI Cambridge 2026.