PM25-Monitoring-Yerevan

Summary

This repository contains a Jupyter Notebook project applying deep learning for virtual PM2.5 sensor estimation in Yerevan, Armenia. It uses data from AirQuality.am (2023-2024) to predict PM2.5 at a held-out monitoring station using measurements from other stations and temporal/environmental features. The project compares four neural network models (MLP, GRU, 1D CNN) with a focus on experimental design to prevent data leakage. Results show a simple MLP performed best, suggesting information availability, not model complexity, is the limiting factor.