By Topic

Forecasting airborne pollen concentration of Poaceae (Grass) and Oleaceae (Olive), using Artificial Neural Networks and Genetic algorithms, in Thessaloniki, Greece

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

4 Author(s)
Voukantsis, D. ; Dept. of Mech. Eng., Aristotle Univ. of Thessaloniki, Thessaloniki, Greece ; Karatzas, K.D. ; Damialis, A. ; Vokou, D.

The impact of airborne pollen on human health was recognized many years ago as high pollen concentrations of specific taxa are responsible for triggering allergic reactions to humans, therefore affecting the quality of life. In this study, we develop data-driven pollen concentration forecasting models for the city of Thessaloniki (Greece), using Artificial Neural Networks - Multi-Layer Perceptron (ANN-MLP). The data correspond to the time period 1987 - 2002 and consist of daily time-series of pollen concentrations and several meteorological parameters. We focus on the taxa of Poaceae (Grass) and Oleaceae (Olive), both known to be of high allergenicity to humans. The input variables (features) for the models were selected with the aid of a multi-objective optimization method that employed genetic algorithms. For this purpose, the number of features and the performance of the models were optimized. The resulting models indicated satisfactory performance with an Index of Agreement (IA) up to 0.93 when predicting pollen concentrations 1 day ahead, whereas the same statistical index decreases to 0.85 when the forecasting horizon is 7 days ahead, meaning that they are suitable for operational implementation.

Published in:

Neural Networks (IJCNN), The 2010 International Joint Conference on

Date of Conference:

18-23 July 2010