Monitoring and predicting of air quality in smart cities using mobile sensor devices

نوع: Type: Thesis

مقطع: Segment: masters

عنوان: Title: Monitoring and predicting of air quality in smart cities using mobile sensor devices

ارائه دهنده: Provider: Anita karim ghassab pour

اساتید راهنما: Supervisors: Dr. Hatam Abdoli

اساتید مشاور: Advisory Professors: Dr. Muharram Mansoorizadeh

اساتید ممتحن یا داور: Examining professors or referees: Dr. Abbasi & Dr.Mohammadi

زمان و تاریخ ارائه: Time and date of presentation: 2025

مکان ارائه: Place of presentation: seminar

چکیده: Abstract: Air pollution is one of the most pressing public-health challenges worldwide. Although fixed air-quality monitoring stations are highly accurate, their capital and operational costs and sparse spatial coverage prevent them from capturing within-city heterogeneity at fine scales. In recent years, low-cost sensors and mobile platforms have emerged as practical complements; however, their raw measurements are strongly affected by humidity, temperature, and enclosure design and are unreliable without data-driven calibration. Building on the design and construction of a mobile sensing device and a complete “measurement–calibration–prediction” pipeline, this study proposes a low-cost, scalable approach for monitoring and forecasting air quality. The proposed hardware is tailored for mobile deployment and enables continuous measurements of particulate matter (PM2.5 and PM10) and other pollutants, along with environmental parameters (temperature and relative humidity) and geolocation coordinates. Data are collected in Hamedan along a predefined route during a specified period. On the data side, we conduct field quality control, spatio-temporal alignment, outlier filtering, and missing-value imputation. We train machine-learning models for calibrating low-cost sensors, including linear regression and tree-based models (random forest and gradient boosting, including XGBoost). The same framework is then used for prediction to estimate pollutant concentrations at selected locations in the city. Model performance is evaluated with standard metrics (MAE, RMSE, MSE, and R²) and compared against reference readings. Results show that calibration aided by environmental variables significantly improves measurement accuracy; the R² achieved for particulate-matter calibration with XGBoost is about 0.66–0.72. For prediction, the Lasso model (RMSE = 0.44–3.50) and the XGBoost model (RMSE = 0.94–3.94) deliver the best performance compared with the other machine-learning and time-series models. The main contribution is an integrated, operational solution that provides urban-scale air-quality mapping and forecasting at low cost with acceptable accuracy.