Russian Neural Network to Forecast Allergy Seasons by Region
Model combines plant pollination dynamics with weather forecasts.
Russian researchers have developed the country’s first domestically built program to help combat seasonal pollen allergies. The project involved specialists from Perm National Research Polytechnic University, the Higher School of Economics, and the Perm State Pharmaceutical Academy.
Until now, Russia did not have its own platform based on local data. The newly presented system analyzes weather forecasts for specific locations in real time along with current airborne pollen concentrations, using a neural network model.
According to Konstantin Shvartz, professor at Perm National Research Polytechnic University and Doctor of Physical and Mathematical Sciences, the algorithm uses these inputs to project peak concentrations for each allergen. This makes it possible to anticipate surges in allergy cases and forecast demand for antihistamine medications in specific regions.
Acting in Advance
The model was trained on data collected over 10 years of observations. Researchers used specialized traps to record daily pollen levels in the air, then manually counted and identified pollen grains under a microscope. They identified nine primary allergenic plants: birch, alder, grasses, maple, elm, pine, poplar, nettle, and ragweed. The algorithm was later adapted using data on pharmaceutical supplies to pharmacies.
By combining these two data streams, the system demonstrated how weather conditions influence pollen release and how, with a delay of several days, peaks translate into increased demand for specific medications. The model enables advance planning of procurement volumes to prevent drug shortages during periods of heightened incidence. Its accuracy reaches 92 percent.
